Kuro5hin.org: technology and culture, from the trenches
create account | help/FAQ | contact | links | search | IRC | site news
[ Everything | Diaries | Technology | Science | Culture | Politics | Media | News | Internet | Op-Ed | Fiction | Meta | MLP ]
We need your support: buy an ad | premium membership

[P]
Sun's rise and fall

By slashcart in Technology
Tue Aug 19, 2003 at 10:37:55 AM EST
Tags: Technology (all tags)
Technology

Sun began the 1990s in an unenviable position. Back then, Sun was mainly a manufacturer of Unix workstations for scientific and technical computing. This market was both crowded, and widely expected to decline as workstation users migrated away from Unix to cheaper commodity machines. It's surprising, therefore, that within 5 years Sun would be catapulted to become the most successful company in the enterprise computing space, multiplying its value by 30 times, and surpassing its formerly much larger rivals like HP. It's also surprising that, within a few more years, Sun would decline just as rapidly as it had risen. How could a middling manufacturer of workstations so rapidly become the darling of IT, only then to become the orphan stepchild? Answering that question is the purpose of this essay.


UNIX THREATENED BY MICROSOFT AND INTEL

Sun's spectacular rise and descent was due largely to how Sun responded to changing market conditions. Understanding those market conditions is therefore necessary for understanding how Sun's fortune could have altered so dramatically.

In the mid-1990s, before Sun's amazing ascent, the entire Unix market was seen as gravely endangered by threats from below. The Intel/Microsoft juggernaut had recently finished destroying its competition in the desktop space, and began eyeing the enterprise market for takeover. This takeover was to have two parts. First, Intel would introduce a new high-end processor to compete against the RISC processors common in the enterprise space. Second, Microsoft would start pushing its new version of Windows (called "NT") as an alternative to expensive and proprietary Unix offerings.

The first part of the takeover, Intel's new processor, would employ a fundamentally new architecture that would be superior to the RISC designs used by the high-end Unix vendors. The main idea behind this new architecture (called "EPIC") was: incoming instructions would be grouped into bundles, and the compiler would decide (at compile time) which instructions would be executed in parallel and how they would be scheduled. This was an extension of the RISC philosophy of shifting work from the processor to the compiler. Since the compiler could spend hours analyzing the instruction stream, it could extract far more parallelism than the RISC processors could extract on-the-fly. And since the logic for determining instruction scheduling and parallelism was no longer embedded in the processor, but in the compiler instead, a huge number of transistors could be liberated for performance-enhancing features like more units, bigger caches, and so on. Thus, EPIC promised to be a revolution as great as the original RISC revolution; EPIC promised vastly greater performance, on smaller die sizes, and for less money.

The software side of the takeover was supplied by Microsoft, in its new operating system called "Windows NT." This new operating system was based on a modern, pre-emptively multitasking, protected memory kernel -- just like the kernels contained in unix operating systems. Thus, it would provide all the reliability and stability of Unix OSes. But NT, unlike Unix, would be common, inexpensive, and easy to use, greatly reducing the administrative expenses and total cost of ownership for organizations that deployed it.

The combination of Microsoft's new operating system, and Intel's new processor, would constitute a crushing blow to the Unix vendors. Combined, Microsoft and Intel would bring high volume, low prices, common components, and superior performance to the weak, fractured, low-volume commercial Unix market. The Microsoft/Intel juggernaut would roll over the Unix vendors, in the same way it did Apple and all its desktop competitors. It became common wisdom, in the mid-1990s, that RISC and Unix were doomed.

This sentiment of doom was repeated by the industry analysts. The analysts agreed that the Intel/Microsoft juggernaut was unstoppable and would bulldoze over proprietary Unix offerings. Soon, the analysts said, Unix would start gradually fading away, and RISC would be destroyed. One very prominent analyst wrote, in 1998, a report on EPIC versus its RISC competitors, and concluded that if the market were a chess match then checkmate had already been declared and EPIC had already won. This was surprising, given that no EPIC processor had yet been released and none was going to be released for three more years. But the other analysts agreed. According to the "common wisdom" of the day, the Microsoft/Intel assault was as good as won, even before it had begun in earnest.




THE UNIX VENDORS RESPOND TO THE MICROSOFT/INTEL THREAT

For the most part, the commercial Unix market agreed with the consensus that it was doomed. Most of the commercial Unix vendors quaked in terror at the prospect of a Microsoft/Intel onslaught, and several Unix vendors reacted by announcing the demise of their own platforms, thereby shooting themselves in the head before Micorosoft and Intel could get to them. SGI announced (in 1997) that it would cancel its upcoming CPUs, cease development of its own processor family, and transition its users to EPIC. HP announced that it would cease development of its PA/RISC processor family and would transition its users to EPIC as well. IBM decided to "hedge its bets" by deploying both EPIC systems and its own POWER systems concurrently, but downplaying the importance of its own POWER systems. And the CEOs of both SGI and HP seriously insinuated that they might cancel their respective Unix OSes (in the distant future) and transition their users gradually to Windows NT.

Sun was the only company which didn't participate in this killing of its own children. Instead, Sun went in the exact opposite direction. Instead of announcing the imminent demise of its processor family, Sun announced that it was designing the next generation of those processors. Instead of insinuating that the Solaris OS might be cancelled, and its users eventually migrated to NT, Sun aggressively promoted the Solaris operating system and derided NT. And although Sun initially dabbled with the idea of building EPIC boxes, alongside its traditional RISC boxes, Sun quickly abandoned that idea and decided to commit all its resources entirely to its own platform. Thus, Sun redoubled its efforts, introduced new models, and even went so far as to purchase (from SGI/Cray) the designs to a massive new Unix server that would be the largest and most expensive Unix server ever. If Unix was dead, Sun apparently was unwilling to believe it.

Because of this unswerving commitment, Sun quickly captured most of the Unix market from its rivals. This wasn't due to technological superiority, but to the fact that Sun appeared to be the only vendor not abandoning Unix. Therefore, Sun won the Unix market because the other unix vendors, believing the market to be doomed, essentially abandoned it. The other unix vendors (in essence) voluntarily ceded the market to Sun.




THE INTEL/MICROSOFT JUGGERNAUT STALLS

The analysts had been in agreement about the imminent demise of Unix. But the difficulty with analysts' predictions is that they're generally either obvious ("computers will get faster next year") or completely wrong. The more analysts agree on some prediction, the more likely they are to be wrong. In predicting the death of commercial Unix/RISC systems, the analysts were virtually unanimous, and the analysts were almost totally wrong.

Cracks started to appear in the Intel side of the takeover when its upcoming Merced processor was dreadfully late, and started falling further and further behind schedule. Schedule slippage was soon accompanied by warnings, issued by Intel, that the new processor's performance would be well below what had been promised. Intel began downplaying the importance of the new processor, stating openly that the subsequent processor ("McKinley") would really be much better. When the new processor was eventually released, its performance, far from being revolutionary, was significantly below currently available RISC offerings on commercial workloads. This tepid performance, combined with a lack of available software, led to disastrous marketplace results. After Merced had been selling for more than a year, sales figures indicated that only single-digit thousands had been sold. Intel's terrifying onslaught had failed to take more than 0.1% of the server processor market.

Microsoft's portion of the assault, the "Windows NT" operating system, was enormous, complicated, and far too immature for enterprise computing. The combination of immaturity and compexity led to innumerable bugs, which Microsoft was incapable of repairing in a timely manner. Soon the buggy OS gained a reputation for serious unreliability, making it inadequate for anything but low-end non-mission-critical servers. Microsoft's marketing machine only exacerbated this problem by convincing credulous managers to convert to NT when NT was not ready, leading to disastrous failures and "re-conversions" back to Unix.

Thus, Intel's EPIC had failed completely, and Microsoft's NT had penetrated only the fringes of enterprise computing. The analysts were proved wrong. The Juggernaut had stalled. The initial foray into the enterprise market had failed.




UNIX EXPERIENCES A RENAISSANCE, TECH ECONOMY EXPLODES, SUN BLOOMS

The failure of the Microsoft/Intel assault left Unix without any serious contenders in the enterprise space. This caused, starting in 1999, something like a Renaissance of Unix. All of the sudden, Unix was seen as being cool again, even "cutting edge," which was surprising for a technology that was 25 years old. Instead of converting away from Unix, as had been so widely predicted, companies began converting to it. And since Sun had picked up so much of the Unix market as its competitors abandoned it, Sun became the natural beneficiary of the Unix Renaissance. Sun's sales and earnings surged.

Adding to Sun's fortune was the earlier release (in 1997) of its new server, the E10000, codenamed "Sunfire." Based on a design Sun had purchased from SGI (which had in turn acquired the design from Cray), this server had 64 tightly coupled CPUs in a huge SMP configuration that was larger and more expensive than any previous Unix machine. When the E10000 was announced, and Unix was presumed dead, a prominent analyst predicted that Sun would sell, at most, a "few dozen" of them. Instead it sold briskly and differentiated Sun from its competitors, since nobody else had anything like it. Sun's E10000 took the top ranking in the heavily followed tpc-c benchmark, being the first server to top 100,000 transactions/second.

But the real boon to Sun was the massive explosion of the tech economy. The Internet suddenly took off like a rocket, and following the Internet boom was the greatest speculative investment boom in the history of mankind. Trillions of dollars flowed into "dot coms" and other technology startups, and all these newfound companies, especially the internet-related companies, required servers. Sun, being the dominant maker of the now well-regarded Unix, seemed like the natural choice. Sun became regarded as the "standard" choice in servers, and as small companies and startups frittered away trillions of dollars of investor money, Sun happily sold them large amounts of expensive equipment. Several companies bought Sun's enormous multimillion-dollar E10000s as a matter of pride and "showing off."

This tech boom, combined with Sun's recent dominance in the Unix market, let to a skyrocketing of Sun's sales and earnings. Sun's stock price jumped to become 30 times higher than in had been. Sun became a $200 billion company that was twice as large as its formerly much larger foe, HP, and that was approaching IBM in market capitalization. Sun was seen as the vendor supplying computers to power the New Digital Age. Sun was on top of the world.




SUN's DECLINE

Thus, Sun's fantastic ascent was largely due to three factors: first, the abandonment of the Unix market by Sun's competitors; second, the technological failures of the Intel/Microsoft juggernaut; and third, the Renaissance of Unix with an accompanying tech economy boom. Each of these factors contributed greatly to Sun's ascent, but each of these factors was, in the end, transient.

The first factor to give way was the tech economy boom. Speculative booms have a history of collapsing as dramatically as they had risen, and the "dot com" boom was no exception. When the collapse came, the NASDAQ stock market lost more than 75% of its value, and almost all of the previously high-flying "dot coms" quickly were driven into bankruptcy. This collapse of the "dot coms" was a special difficulty for Sun, since almost all of them had been Sun customers, and Sun was heavily reliant upon them for revenue. So, as the dot coms went bankrupt, the market for Sun's equipment sharply contracted. This contraction was exacerbated by the fact that these bankrupt startups dumped piles of their mostly unused Sun hardware back onto the hardware market, at a fraction of the original price, creating a glut of Sun equipment.

But the tech economy collapse was to be only the first of Sun's problems. Micorosft's and Intel's initial forays into the enterprise space might have been disasters, but those two companies are endlessly tenacious and have almost unlimited resources to devote to new attempts. So, of course, they would try again. Microsoft released a new version of its NT operating system (now renamed "Windows 2000"), and the new version proved to be vastly superior to the old. And Intel released a new version of its Itanium processor (called "Itanium 2"), and its strong performance surprised almost everyone, offering benchmark scores similar to the best of the RISC processors. Thus, the assault from below, though it had suffered some initial setbacks, seemed once again poised to encroach on the commercial Unix market. And this time, the threat from below had an unexpected new component. The upstart Linux operating system burst into view and offered Unix reliability, not just for less money, but for free. Even though Microsoft's OS was still seen as having problems, Linux was regarded as highly reliable and absolutely adequate for low-end server tasks.

Further contributing to Sun's decline, was HP's and IBM's "re-commitment" to Unix. HP and IBM both recongized their enormous strategic blunder in cedeing the Unix market to Sun, and both companies sharply reversed course and began dedicating themselves to re-taking lost Unix marketshare. Both companies aggressively designed and promoted new machines to compete with Sun's. Both designed massive high-end SMP machines, HP's Superdome and IBM's S/80 (later, IBM's p690), to compete with Sun's E10000 and upcoming E15000. And both companies loudly proclaimed their sudden loyalty to Unix. Since both IBM and HP have more capital and more system-designing experience than Sun, both their high-end servers edged out Sun's in performance.

Thus, Sun was assaulted on three fronts. Sun's low-end marketshare was eroded by Linux and Windows 2000 running on commodity servers. Sun's high-end marketshare was eroded by new, more powerful servers from the freshly re-committed HP and IBM. And Sun's principal customers, "dot coms" and technology startups, disappeared as the tech bubble popped. These three factors were disastrous for Sun. Sun's sales and earnings plumetted; Sun's stock lost over 90% of its value; and Sun retreated back to the same position it occupied in the early 1990s, as a middling maker of Unix machines in a crowded market.




SUN's PROSPECTS

Sun's recent difficulties have led some to predict that Sun will fade into oblivion, going the same route as Digital Equipment. This danger may be exaggerated. Sun is financially well-managed and isn't in any kind of short-term crisis. But Sun faces serious long-term problems.

One of the most significant problems is the growing gap in performance between Sun's UltraSparc line of processors, and the processors offered by Intel or IBM. For a long time Sun's UltraSparc processor line had marginally trailed its competition in performance, but with the UltraSparc-III, introduced in 2001, the gap in performance considerably widened. This was due to two factors. First, Sun's CPU development team is less effective than the advanced and well-funded teams of Intel or IBM. Second, the process technology used to manufacture Sun's CPUs (by TI) is older than that found in the extremely expensive cutting-edge fabs of IBM and Intel. These two factors combined led to the UltraSparc-III CPU that is in-order, has longer pipelines, has fewer units, and runs at a lower clock rate than its competitors. The poor performance of the UltraSparc-III led some industry watchers to speculate that Sun is now incapable of designing and manufacturing high-end CPUs that are competitive with those from IBM or Intel. Since CPU design and manufacture are becoming even more capital-intensive every year, and Sun has far less capital than either Intel or IBM (and Sun's fab partner is willing to devote less of it), it's presumed that Sun will be unable to narrow this performance gap in the future.

Sun's response to this threat is to use an entirely new approach to CPU design, called "throughput computing." Sun claims that throughput computing will represent a revolution in CPU design and will allow chips that are 30x faster than those of today. The idea behind "throughput computing" is to replace the traditional massive, complicated processor core with a large number (>8) of small, extremely simple processor cores, each of which is capable of executing its own thread. Thus, a "throughput" CPU is actually a horde of tiny processors acting in tandem. The advantage of this approach is that it avoids the rapidly diminishing returns of expending exponentially more processor logic to extract smaller and smaller amounts of per-thread parallelism. Of course, there's a downside: single-thread performance suffers. But the tradeoff of single-thread performance for total instruction throughput is highly desirable for some applications. To most server applications, single-thread performance is unimportant, and total instruction throughput is all that matters. For these applications, Sun's new "throughput" CPUs might offer far better performance for a comparable die size and manufacturing process.

The difficulty is that Sun is more than one full process generation behind in manufacturing process, and it's unlikely that Sun's fab partner, Texas Instruments, is going to invest in new fabrication equipment on Sun's behalf. It's unknown whether Sun's "throughput computing" approach will offer significant enough performance benefits to compensate for older manufacturing techniques. Barring revolutionary advances in performance from the throughput apporach, the most Sun could hope for is to equal the performance of its competitors on server applications. Even if Sun's "throughput computing" proves to be a superior approach †for processor design, this advantage will likely be short-lived since there's nothing preventing IBM or Intel from employing the same ideas (Sun does not own the concept of SMT).




WILL THE SUN RISE AGAIN?

Thus, we come to the final question: Will Sun rise again? The three factors underlying Sun's prevoius ascent are permanently gone. The tech economy won't likely boom again like it did in the late 1990s; Sun's competitors won't likely try to kill off their own platforms; and Intel and Microsoft won't likely release entirely novel, untested versions of NT or Itanium. So, a repeat of the ascent similar to the one in the late 1990s is unlikely.

Furthermore, it's unlikely that Sun's low-end boxes will be able to compete on price/performance with the offerings from research-lean organizations like Dell. And it's unlikely that Sun's high-end boxes will be able to compete with offerings from massive, well-funded, technologically sophisticated companies like IBM and HP.

These facts have led some industry watchers to predict that in the long term, Sun is doomed. But if there's any lesson to be learned from Sun's history, it's this: extrapolating from current trends is not a good way of predicting the future. Sun is a master of pulling tricks out of its hat. Sun has been written off before, and it prevailed.

Sponsors

Voxel dot net
o Managed Hosting
o VoxCAST Content Delivery
o Raw Infrastructure

Login

Poll
What's in store for Sun?
o Doomed to bankruptcy 1%
o Will fade away gradually and be bought out 38%
o Will remain modestly successful 44%
o Will gain marketshare and revenues 13%
o Will regain former dot-com glory days 1%

Votes: 72
Results | Other Polls

Related Links
o Also by slashcart


Display: Sort:
Sun's rise and fall | 159 comments (134 topical, 25 editorial, 0 hidden)
Premature Wintel, Lintel Propaganda :::peniz:::Q (2.55 / 9) (#2)
by peniz Q on Mon Aug 18, 2003 at 10:08:13 PM EST

Re-section to Op-Ed.

This is more wishful thinking than astute analysis. The first mistake is that you attribute strategic directions of the companies involved to over-generalized forces.

One example is that you tout the "unreliability" of Windows NT as the reason it was not able to gain in the high-end market. Your Linux bias is clearly evident in that statement. NT was never designed to be a high-end server, much less a low-end one. It was designed to be a stable workstation, which it was. NT was never a 64-bit OS. How was it ever supposed to be a high-end server?

Another example of your flawed logic is the fact that you never pay any lip service to Sun's forecasted road map. You neither mention the new Sparc chips on their way, nor the Mad Hatter Linux workstation project soon to be unveiled.

Sun has nothing to prove to "analysts". It only needs to make a nice profit like Apple. Apple only has about 3% of the market share, but it's products are the best in the computer industry. Sun needs only to follow that path.



Dangerous content.

pro-Linux bias? (4.00 / 2) (#9)
by Ugly on Mon Aug 18, 2003 at 10:53:43 PM EST

I was thinking the opposite.

If anything, the article doesn't spend enough time on the threat that Linux poses to Sun. It's a far greater threat than Microsoft is. Windows is getting to be sufficiently reliable for server use, but it will never be as easy to port existing software from proprietary Unix to XP as to Linux. Also, Microsoft killed off NT running on non-x86 platforms. Linux runs on much of the big iron hardware.



[ Parent ]
Linux and Big Iron (4.33 / 3) (#40)
by Metatone on Tue Aug 19, 2003 at 10:20:38 AM EST

Indeed, not to beat the Linux drum too much, or hijack a story about Sun, but Linux certainly presents a serious difficulty for Sun. As the article states, in principal, SGI, HP and IBM originally capitulated from the market, promising a move to EPIC and curtailing development of their UNIX OS's somewhat. (SGI and HP did, IBM wavered.) The rise of Linux gave each of these companies an opportunity to get back on the UNIX trail and very importantly holds out the possibility of solving the old "it's UNIX but it won't run your app without a serious porting session" problem.

In the bad old days UNIX became seriously fragmented, which was good for holding onto current customers (hard to move to a different platform) but bad for acquiring new ones. By promoting Linux for Big Iron, HP, SGI and IBM hope to establish a standard Unix-like OS to sell their hardware and consulting solutions around. Whether this strategy beats out the integration of Solaris and Sun hardware remains to be seen. However, Oracle is appearing on Linux installations here and there (which was certainly one of Sun's original home territories) and suddenly Sun seems to have fewer and fewer distinct advantages over IBM, HP and SGI, especially since they can argue (for example) "Solaris will lock you in to Sun, use Linux on SGI and if you ever need to, you can switch to IBM or HP easily... Sun can have you over a barrel, and you know how much that hurts..."

Anyhow, as the OP notes, Linux is a more direct threat to Sun than NT as porting is much less complex. It also promises enterprise wide functionality which has it's attractions over being a split Sun/MS shop.

One final thought, don't ignore the rise of Apple's OSX in the UNIX workstation realm. As good as Lintel (Linux/Intel) solutions have been in taking workstation sales away from Sun, Apple has the potential to steal even more. Many places have acquired Sun workstations as an integrated hardware/OS package, a niche Apple would seem to compete in better than Linux, at least for now. Workstations might not make as much as servers for Sun, but losing mindshare can be fatal. Not only can it decrease the "admin gene pool" but if the workstations stop being Sun, the server is psychologically easier to replace too.

[ Parent ]

msft (4.50 / 4) (#23)
by slashcart on Tue Aug 19, 2003 at 02:01:37 AM EST

NT was never designed to be a high-end server, much less a low-end one. It was designed to be a stable workstation, which it was. NT was never a 64-bit OS. How was it ever supposed to be a high-end server?

At that time Microsoft was heavily promoting NT as a replacement for Unix in the enterprise. Microsoft put out a bunch of different advertisments of enterprise customers that had switched to NT (the Chicago Board of Options Exchange was one of them). Sun countered with a number of ads of its own; one said "considering switching your enterprise apps to NT?" then it showed a boxer in a ring getting the crap kicked out of him, then it said "isn't there enough suffering in the world?"

One example is that you tout the "unreliability" of Windows NT as the reason it was not able to gain in the high-end market. Your Linux bias is clearly evident in that statement.

The unreliability of Windows NT 4.0 was widely acknowledged. Even people who believed that NT was going to dominate the enterprise space believed that NT 4.0 had reliability problems. Dvorak wrote in his column (back then): "With Windows NT approaching 23/6 reliability, who can stand in its way?" or something to that effect.

Believe me, I don't have a Linux bias.

[ Parent ]

You've been smoking too much pot (4.66 / 6) (#33)
by jungleboogie on Tue Aug 19, 2003 at 05:40:56 AM EST

Your "wishful thinking" shows complete ignorance to reality.  The first mistake is that you don't know what the hell you are talking about.

One example is that you are trying to explain how Windows NT was never meant to "be a high-end server, much less a low-end one".  Please explain why Microsoft distributed it with database, e-mail, file, and web server applications? Please explain why there are "server" and "workstation" versions? Microsoft has been trying for years to push it into the high-end, but the fact of the matter has always been that, especially during the days of NT 3 and 4, it was not considered by anyone to be an "enterprise" solution.

Another example of your flawed logic follows in "NT was never a 64-bit OS".  I guess the versions of Windows NT and Windows 2000 for the 64-bit DEC Alpha chip is just a nightmare that I keep having!  The 64-bit versions of SQL server.  Oh, the horror!!  If you seriously think that you know anything about the history of any high-end computing, or even of Microsoft Windows for that matter, you should save it for your buddies on the playground. Otherwise, try and _listen_, you might actually learn something.

As for the rest of your comments, I find them to be of dubious value, especially considering your badly flawed perception on such basic material facts of this industry.

[ Parent ]

I agree mostly... (none / 0) (#81)
by styrotech on Tue Aug 19, 2003 at 07:19:08 PM EST

Another example of your flawed logic follows in "NT was never a 64-bit OS". I guess the versions of Windows NT and Windows 2000 for the 64-bit DEC Alpha chip is just a nightmare that I keep having!

But the (released at least) Alpha versions of NT were 32bit operating systems.

[ Parent ]

What? (none / 0) (#93)
by jungleboogie on Wed Aug 20, 2003 at 01:31:25 AM EST

With 64 bit pointers, 64 bit longs, the Alpha is a 64 bit platform.  How could Windows NT running on it be a "32-bit operating system"?

If it uses the 64 bit registers, and 64 bit values, what exactly are you talking about?

[ Parent ]

Nope (none / 0) (#104)
by komadori on Wed Aug 20, 2003 at 09:39:27 AM EST

The Alpha may be a 64-bit processor, but WNT/Alpha was not a 64-bit operating system. The programming model was ILP32 and the virtual address space was limited to 4gb. Special PALcode (horizontal microcode) in the ARC console firmware made the MMU behave in a manner similar that of a MIPS32 or IA32 processor in order to make WNT easier to port.

"When we are victorious on a world scale I think we shall use gold for the purpose of building public lavatories in the streets of some of the largest cities of the world." Vladimir Illich Lenin


[ Parent ]
Interesting (none / 0) (#112)
by jungleboogie on Wed Aug 20, 2003 at 12:21:49 PM EST

I guess this explains why Microsoft is taking so long for their amd64 64-bit OS. Being Microsoft, I imagine they are running into thousands of bugs or design errors that they didn't see in a 32bit model.

[ Parent ]
History (none / 0) (#86)
by truchisoft on Tue Aug 19, 2003 at 08:46:41 PM EST

hmm, i've got the knowledge worm after reading this article i know a google search will get me there, but i simply dont have the patience to scroll 20 pages before finding anything about the topic. Would you be so kind as to point me somewhere to read about all this stuff?? (any history webpage or something?) - Down with Trolls!... Long live Goblins!
--- Saludos de Argentina.
[ Parent ]
Sun hardware (4.00 / 1) (#7)
by komadori on Mon Aug 18, 2003 at 10:45:48 PM EST

I thought Sun was going to go somewhere with MAJC but it's appears to be pretty much dead now. They've removed it from their microelectronics website.

Still, even with Sun's current problems with their hardware lineup, they have Solaris and it's king among operating systems. I just wish they would concentrate more on Solaris/IA32 and less on Linux. The LX50 actually ships with Linux by default... -_-

"When we are victorious on a world scale I think we shall use gold for the purpose of building public lavatories in the streets of some of the largest cities of the world." Vladimir Illich Lenin


Solaris King. NOT! (2.50 / 2) (#65)
by johnnyfever on Tue Aug 19, 2003 at 01:16:13 PM EST

We have several E10Ks at work here as well as numerous smaller Sun servers. We also have several AIX boxes. Without exception, the Unix admins far prefer AIX due to it's far superior (read: existant) administration tools. These are guys who have been doing this for ~15 years.

[ Parent ]
Your admins must be... (4.00 / 1) (#76)
by FieryTaco on Tue Aug 19, 2003 at 03:33:49 PM EST

They must be MCSEs from before there were MCSEs. Most competent admins I know do not prefer AIX, nor do they find high level administration tools the prefered method of running a box.

[ Parent ]
"Whip me, beat me, make me maintain AIX" (3.00 / 1) (#80)
by komadori on Tue Aug 19, 2003 at 06:22:12 PM EST

They're probably just addicted to smit(ty) ^_^.

"When we are victorious on a world scale I think we shall use gold for the purpose of building public lavatories in the streets of some of the largest cities of the world." Vladimir Illich Lenin


[ Parent ]
That's nice. (none / 0) (#90)
by MKalus on Wed Aug 20, 2003 at 12:05:42 AM EST

For the past 12 years or so *I* adminstrated Unix Systems myself, Sun, HP, and (shudder) SCO as well as AIX.

Out from all these I came to appreciate Solaris the most. Why? Because it is still true to the Unix "spirit" so to speak.

Sure, I don't get my nifty GUI utilities like in HP and AIX, I also don't have to compile my Kernel new evertime I make a change (thanks SCO), but in the end I can do one thing with Solaris that I really like: I can make the system as bare bone or full blown as *I* want it to be, without loosing anything.

Try that with AIX or HP-UX, most of the admins I know who administrate HP-UX for a living have a hard time to do anything if they don't get their GUI tools.

Sad.
-- Michael
[ Parent ]

Really!?! (none / 0) (#103)
by porkchop_d_clown on Wed Aug 20, 2003 at 09:25:57 AM EST

My AIX experience is 6-7 years old, but I remember it as being must more frustrating than Solaris.

I guess things have gotten better!


--
You can't raise my prices. You can't build more power plants. You can't build more power lines. Why are my lights out!?!


[ Parent ]
Then they're not particularly good admins (none / 0) (#110)
by irixguy on Wed Aug 20, 2003 at 12:13:18 PM EST

Out of those 3, Solaris is by far and away the most powerful, and the easiest to administer.

By a long shot.

Sun have focussed on giving tools to admins - IBM and HP have focussed on shiny! tools to try and help their sell - trinkets that impress the CTO during demo time.

AIX is cumbersome and awkward to manage, and it lacks functional command line tools. Anyone who relies on SMIT or SMITTY is a fool who doesn't know their system. HP-UX lacks lots of low level admin tools - it's positively barren in that regard.

Solaris is simple and easy to manage and to script from the command line, and it has a lot of tools to make your life as an admin easier.

I would comment on how you are making comments on behalf of your admins, but I shall ignore that .....

I'm making these comments as someone who has adminned all three OSs for the last 10 years.


--

"Tell them we are not Gods, but Sysadmins, which is the next best thing."


[ Parent ]
Interesting (2.50 / 2) (#124)
by johnnyfever on Wed Aug 20, 2003 at 04:42:53 PM EST

Wow this comment sure has brought out the brand loyalties of people and caused some very entertaining knee jerk reactions and name calling!

You can go ahead and "comment on how I'm making comments on behalf of my admins", I'm simply repeating what they have told me directly. I know both of these distiguished gentlemen quite well, one of whom I have worked with for 6 years, the other for three, and I call both of them friends outside of work.

As I mentioned in the original comment, both of these fellows have been admins for ~15 years, and that's being conservative. They have looked after lots of different OSs in their time.

I'm all for simplicity and so on, that's why I personally use slackware at home. Not being an admin myself, I am not qualified to say how useful or useless things like smit are, I'm taking the word of two very experienced people who have been the best admins I have worked with. The suggestion that these fellows 'rely' on the 'shiny!' tools due to their lack of knowledge about their systems is beyond absurd.

From my own experience as a developer, I can say that both the Sun boxes and the AIX boxes we use are mostly quite reliable, however we have had more unplanned outages on the Sun boxes. A couple times it has been due to OS bugs which we required Solaris patches to fix .. fair enough, that happens to everyone. What I have found un-excusable is the incredibly shitty hardware that the Sun boxes seem to be made of. We have 5 E10Ks here and the number of disks that we have lost is just stupid. Especially considering the outrageous price Sun charges for a 'brick' of disks, or whatever they call them. I could by a bigger and far more reliable drive from a second hand PC store! All in all the E10Ks have been very lemonish, and the Sun contractors don't seem to have the first clue about what's going on most of the time. Also, the idea of dividing the e10ks up into domains is great and all, except when you have to take the whole frame down, affecting far more people than you would with a couple of independant and far cheaper servers. Also, IBMs HACMP has performed flawlessly for us on the odd occasion we've needed it ... there doesn't seem to be a comparative equivalent for Solaris.

I guess my point is, that from a developer's perspective, I have not been any more impressed with Solaris on Sun hardware than I have been with any other Unix-variant. If anything, I have been less impressed by Sun. The two most experienced sys-admins I know have not been particularly impressed with Solaris either. I just don't get what's supposed to be so fucking great about it, and why people like yourself seem to feel compelled to defend it to the end. There is nothing special about it in comparison to it's competitors.

Flame on.

[ Parent ]

(OT) Why??? (none / 0) (#136)
by BJH on Thu Aug 21, 2003 at 03:02:28 AM EST

Why do you have that as your sig? It's not even correct. You've got 'ookiku' (an adverb) modifying 'sakana' (a noun), not to mention that 'shitsuhou' isn't a real word.

Are you deliberately trying to yourself?
--
Roses are red, violets are blue.
I'm schizophrenic, and so am I.
-- Oscar Levant

[ Parent ]

Er... (none / 0) (#137)
by BJH on Thu Aug 21, 2003 at 03:04:01 AM EST

Last line was meant to be "Are you deliberately trying to embarrass yourself?"

Um.
--
Roses are red, violets are blue.
I'm schizophrenic, and so am I.
-- Oscar Levant

[ Parent ]

leaving out things (none / 0) (#15)
by turmeric on Tue Aug 19, 2003 at 12:06:34 AM EST

BSD.

who are sun's customers?



BSD is dying... get with the times... /nt (1.66 / 6) (#16)
by rmg on Tue Aug 19, 2003 at 12:08:16 AM EST



_____ intellectual tiddlywinks
[ Parent ]

All you need to know about Sun in just 12 words (4.00 / 5) (#18)
by ghjm on Tue Aug 19, 2003 at 12:13:30 AM EST

"The Network Is The Computer."  Yep.
"The Dot In Dot Com."  Nope.

-Graham

This is like History class... (2.44 / 9) (#20)
by Dinner Is Served on Tue Aug 19, 2003 at 01:05:33 AM EST

except it's Unix History. Who the hell would find that interesting? This is why TechTV is off the air.
--
While I appreciate being able to defend against would-be rapists who might suddenly drop in from the sky, I don't appreciate not being able to see the Northern Lights. -- mfk
Hrm... (4.00 / 3) (#24)
by GavalinB on Tue Aug 19, 2003 at 02:04:18 AM EST

Isn't TechTV on its own channel now?
---
The Future is Prologue: Join Our Sagas Today!
[ Parent ]
Answering my own question... (4.00 / 3) (#25)
by GavalinB on Tue Aug 19, 2003 at 02:06:14 AM EST

Yes.
---
The Future is Prologue: Join Our Sagas Today!
[ Parent ]
Another problem with Sun (2.00 / 3) (#28)
by ZorbaTHut on Tue Aug 19, 2003 at 02:16:24 AM EST

There's another thing Sun has been doing that's almost suicidal, as near as I can tell. They're practically trying to drive themselves out of business.

You see, Sun has made this amazing new programming language, called Java. Java does many neat things. Java runs on Sun computers quite well. However, another thing Java does is run on *any* computer. It's worth pointing out that Sun is basically giving Java away for free.

So: think about this for a second. You have a hardware company releasing a product, for free, that lets people write good efficient code for *any platform*. Including, say, relatively inexpensive Intel computers.

I don't get what they're trying to do here. Economically it makes no sense at all. You don't make it *easier* for the market to get rid of your product - you make it easier for them to *use* your product, yes, but Java is no easier on Sun computers than it is on any others.

the other side of the coin (5.00 / 6) (#31)
by khallow on Tue Aug 19, 2003 at 04:03:40 AM EST

You see, Sun has made this amazing new programming language, called Java. Java does many neat things. Java runs on Sun computers quite well. However, another thing Java does is run on *any* computer. It's worth pointing out that Sun is basically giving Java away for free.

I think it's worth pointing out that Java wouldn't be adopted if it weren't free. And it would be worthless if it ran only on Sun's machines. The real problem is that Sun's hardware and their Solaris OS no longer has a compelling future. Too many people work on cheap platforms like Windows/intel or linux/intel. I eventually foresee all that Sun hardware going away. Then the question is what do they have that will survive more than say five years? It's their products built on the Java platform and similar things.

Stating the obvious since 1969.
[ Parent ]

wrong (none / 0) (#53)
by asad on Tue Aug 19, 2003 at 12:27:05 PM EST

you have to consider the work that thos people are doing on the cheap platforms.  Sun has way too big of a customer base and too much cash to go away in the next 10 years nevermind 5.  Even in the worst case scenario look at SGI they are still a walking zombie.  Their idea of grid computing has started to take hold in some chip places.  here's an article from the register that gives more details.

[ Parent ]
Strategy vs. Tactics (4.50 / 2) (#38)
by abulafia on Tue Aug 19, 2003 at 09:15:13 AM EST

Owning a platform to which people write applications is much more important than owning a hardware platform.

This is why MS both carefully controls the API for windows and got so freaked out about Java/Netscape, back in the day. Too bad Netscape was such a screwup - they never should have announced what they were attempting.

(Of course, there's a problem; they wouldn't have had such an amazing IPO without the hype. I still think stealth would have worked better - if they'd actually built a decent platform before they became arrogant, they would have had a chance. You don't always need IPO cash to develop. But what do I know, I'm not rich yet.)

[ Parent ]

It makes perfect sense (5.00 / 1) (#46)
by Simon Kinahan on Tue Aug 19, 2003 at 11:13:19 AM EST

Sun see themselves principally as a hardware vendor. They make money selling and supporting hardware. They only write and maintain Solaris because noone else will write OSes for their hardware. However, the value of hardware comes principally from the software it runs.

Most software is not written for Solaris. Even Unix software is rarely written specifically for Solaris. The market is just too small. Most software was (and much still is) written specifically for Windows. Back in the early days of Java, there was a serious risk that writing for or supporting any other platform would be uneconomical (rememer Linux was still just a geeky toy at that time). Apart from the classic Unix apps, which hadn't evolved much in 10 years, and some speciality software, you could have counted the solaris software packages available on the fingers of one hand.

Sun see Java as their solution to this problem, although arguably the rise of Linux and free software has made it less important than it could have been for their survival. Sun have pushed Java into every possible emerging software market, and as a result it has become the language of choice for enterprise application development, and a decent contender in portable software, GUI development and several other fields. As a result, all that software runs on Solaris. It doesn't matter that it also runs on every other platform, because if Java *only* ran on Solaris, noone would ever have written any Java software.

Simon

If you disagree, post, don't moderate
[ Parent ]

Where do people get this? (5.00 / 1) (#68)
by spectra72 on Tue Aug 19, 2003 at 01:38:33 PM EST

"Sun see themselves principally as a hardware vendor"

I see variations of this theme in every single discussion about Sun, and frankly, from my perspective as a Sun employee, nothing could be further from the truth.

In my insider opinion, Sun sees themselves as a "Systems" company. The Micro*systems* moniker is quite telling in this regard.

Of course, it goes beyond a simple name, that would be too trite. When one thinks of Sun's contributions to the Unix landscape through the years, what comes to mind? Hardware? Not likely. Try Solaris, NFS and JAVA.

When one looks at where Sun spends huge chunks of its R&D budget every year, do you think it is all spent on Hardware? Not likely. Try Software. (With SPARC chip R&D being way up there as well.)

Now people may say that Sun makes no money on Software, (they'd be wrong to say that, but I'll play along), but that still says nothing about how Sun sees itself.

Sun has a broad product line, from massively scalable "Big Iron" machines to 4-8 way lower end servers down to 1-2 way Intel machines. Practically the entire server/workstation line is capable of running Solaris (sparc and x86). Sun also offers Linux on the low-end machines. Sun sells storage hardware from small 1U disk racks to large, multi-terrabyte DataCenter Class arrays. Sun has JAVA. Sun has webserver/application server/portal server/identity server software suites. Sun has a Service Division that provides support for everything it sells and Professional Services to help customers architect and plan just about any computing need they have. All of this adds up to more than "just a hardware company".

I say this not to defend or refute anything in Sun's history nor to try to spin this story into some sort of Sun wank-fest. But inevitably, someone starts a post with "Sun thinks of themselves as a hardware company..." and then plays armchair MBA, trying to create a business plan for Sun in 50 words or less. Go to it I say, but at least have the correct idea how Sun actually sees itself.

[ Parent ]

Okay, then ... (none / 0) (#72)
by Simon Kinahan on Tue Aug 19, 2003 at 02:49:02 PM EST

Its wrong for me, as someone with no particular insight into Sun, to talk about how the company sees itself, so let me rephrase: Sun makes most of its money by selling server hardware, and a smaller amount of money by selling services, and an almost insignificant amount of money selling software. Like all other companies whose revenue comes from similar sources, they doubtless realise that you don't make money just shifting boxes. You have to provide software and services too. This doesn't change the fact of where the revenue comes from.

Java, strategically, makes perfect sense in that light, as I explained above, as do all the other projects you mention. Which is far from meaning that Sun doesn't develop software or see it as important, or spend large amounts of R&D money on it. Clearly they do. It just isn't where the money comes from.


Simon

If you disagree, post, don't moderate
[ Parent ]

Java over Win32 (5.00 / 1) (#55)
by pbreton on Tue Aug 19, 2003 at 12:32:21 PM EST

Java was jujitsu for Sun.

Develop on Windows, deploy on Unix.

If development was done with the Win32 API, deploying on Unix would not be a (serious) option.

C# has a different problem: Microsoft has to migrate both VB and C++ developers, hence the emphasis on *language* instead of *platform*.



[ Parent ]
simple (none / 0) (#121)
by treat on Wed Aug 20, 2003 at 01:54:00 PM EST

Solaris on the Sparc was supposed to be the preferred platform for Java. Sun even promised coprocessors. Unfortunately, it's slower than both Linux and Windows on an x86 of comparable speed. And on price/performance it loses terribly of course.

Had Sun ever made the Java coprocessor that they originally promised, people might buy Sun machines to run Java software on and Java might be perceived as fast.

[ Parent ]

Good strategy by Sun (5.00 / 1) (#123)
by coljac on Wed Aug 20, 2003 at 03:27:38 PM EST

Java is great technology and a fantastic strategy for Sun. Imagine if there was no J2EE, how much more ground .NET might be making. Imagine how much more difficult it would be for developers to work on cheap boxes and deploy reliably on big production hardware. There would definitely be more of a trend towards Windows for enterprise computing if Java wasn't around. So while it might not be making buckets of money for Sun on the surface, it's certainly keeping their biggest competitors from eroding their market share more than need be.



---
Whether or not life is discovered there I think Jupiter should be declared an enemy planet. - Jack Handey
[ Parent ]

What laughably oblivious nonsense. (3.04 / 24) (#35)
by RobotSlave on Tue Aug 19, 2003 at 06:58:59 AM EST

First, you show absolutely no understanding of Sun prior to the mid-90s. That means you've ignored more than half of the company's history, and that ignorance is quite apparent.

Second, you mention DEC only in passing, as an example of a company that has fallen, and make no mention whatsoever of the substantial role played by the Alpha processor in that brief historical window that you've chosen to describe.

Third, your article makes no mention of a little company called "Oracle," and of the role that that oh-so-insignificant firm might have played in the sales of E10K machines, and in the broader "big iron" market in those few years you've singled out.

In short, you didn't do your homework, and as a result, you've naïvely framed your story in the facile, politically compromised terms of the contemporary server market, instead of delving into the more interesting history that even a cursory bit of research would have revealed.

Congratulations, you're pretentious and stupid (4.33 / 9) (#41)
by slashcart on Tue Aug 19, 2003 at 10:21:00 AM EST

First, you show absolutely no understanding of Sun prior to the mid-90s. That means you've ignored more than half of the company's history, and that ignorance is quite apparent.

The article was about Sun's spectacular rise and fall. Thus the title: "Sun's rise and fall." Sun's history during the 1980s was long before the spectacular rise and fall, and so is not relevant to the article.


you mention DEC only in passing, as an example of a company that has fallen, and make no mention whatsoever of the substantial role played by the Alpha processor in that brief historical window that you've chosen to describe.

The article was not about the DEC Alpha. Nor was it about a "brief historical window" as you claim. Once again, the title of the article: "Sun's rise and fall." The DEC Alpha was not relevant to Sun's rise or fall; it would've been absurd to wander so far from the topic as to start discussing the DEC Alpha. Unless you seriously think that Sun's fall was caused by Sun's customers migrating to DEC Alpha.

You apparently completely misunderstood the title of the article. I don't see how that could've happened.


Third, your article makes no mention of a little company called "Oracle," ... and in the broader "big iron" market

This grows tiring. Once again, "Sun's rise and fall," not "The history of Big Iron" or "Oracle." Or perhaps you think big iron caused Sun's fall? Perhaps the Sun customers were migrating to S/390s?


as a result, you've naÔvely framed your story in the facile, politically compromised terms of the contemporary server market
Here it seems like you're desperately trying to piece together an impressive-sounding sentence by using terms like "naive," "facile," and "politically compromised." All of those terms are trite, vague, and inapplicable to this context. For example, the terms of the server market are "politically compromised?" How are the terms politically compromised?

p.s. You deserve a flame, because your comment was so outrageously childish. I understand completely that you disagreed with the article, but it's possible to voice your disagreement in a more reasonable way.

[ Parent ]

insulting your readership (2.42 / 7) (#47)
by DJ Glock on Tue Aug 19, 2003 at 11:44:57 AM EST

smart move.

maybe after you learn how to write you'll learn how to take criticism.

*** ANONYMIZED ***
[ Parent ]

He is not my "readership" (4.40 / 5) (#51)
by slashcart on Tue Aug 19, 2003 at 12:18:57 PM EST

maybe after you learn how to write you'll learn how to take criticism.

I have nothing against criticism. But he phrased his comment in a most inflammatory way. The article wasn't insulting and didn't provoke that. The article wasn't even that contentious. I didn't insult anyone's favorite technology or operating system.

Note that the rest of the discussion here is completely civil.

[ Parent ]

duh (2.00 / 6) (#52)
by DJ Glock on Tue Aug 19, 2003 at 12:24:30 PM EST

his comment was crafted to provoke a response from you. it's called trolling.

if you can't resist the urge to engage in flame wars, maybe it's time for you to grow up.

*** ANONYMIZED ***
[ Parent ]

That was not criticism... (5.00 / 3) (#63)
by johnnyfever on Tue Aug 19, 2003 at 01:08:32 PM EST

It was a rude, childish and somewhat misplaced troll. It has been my experience that the best way to deal with a heckler is to put them down quickly and soundly. I think slashcart succeeded there with his response.

[ Parent ]
And you, sir, are still oblivious. (3.44 / 9) (#77)
by RobotSlave on Tue Aug 19, 2003 at 03:42:47 PM EST

"The article was about Sun's spectacular rise and fall. Thus the title: "Sun's rise and fall." Sun's history during the 1980s was long before the spectacular rise and fall, and so is not relevant to the article."

Christ, what a dimwitted ass you are. You have most certainly not described the "spectacular rise and fall" of Sun— instead, you've concocted a poorly researched description of things that happened during the spectacular rise and fall of Sun's market cap. More importantly, the explanations you've cobbled together for that "spectacular rise and fall" are utterly laughable when compared to the real explanation— to wit, that this "spectacular rise and fall" you're so fixated on had little to do with Sun, and everything to do with a speculative bubble in tech shares.

Those words I used that got you so riled up most certainly are applicable to the astonishingly bad article you've written.

When I referred to the politically compromised terms of your argument, you stupid oaf, I was not complaining about the words you were using, but rather to your slashdotesque framing of the story in terms of, and I quote, "the Intel/Microsoft juggernaut" vs. everyone else.

Your article isn't about history— no, it's little more than anti-Microsoft, pro-UNIX agitprop, as you've cast aside historical events of enormous significance in the course of shaping the story fit your preconceptions.

My tone is insulting, yes, and quite deliberately so, as I believe propaganda, even ingenuous propaganda, needs to be met with derision delivered in the sharpest terms available.

[ Parent ]

OK, troll.. (4.00 / 3) (#83)
by slashcart on Tue Aug 19, 2003 at 07:51:38 PM EST

You have most certainly not described the "spectacular rise and fall" of Sun .. you've concocted a ... description of ... the spectacular rise and fall of Sun's market cap

The article described the rise and fall of Sun from a technological perspective; market cap was briefly mentioned only twice in the 5 page article. Anyway market cap closely follows an underlying company's success or failure.

When I referred to the politically compromised terms of your argument, you stupid oaf, I was not complaining about the words you were using

I know what you were trying to say.


Your article... [is] little more than anti-Microsoft, pro-UNIX agitprop

There's little or no attempt in the article to persuade readers that Unix is currently better than Microsoft. That troll isn't even plausible.


My tone is insulting .. as I believe propaganda ... needs to be met with derision delivered in the sharpest terms available.
Your tone was childish and inappropriate. I realize that's the point of a troll. The only reason I responded was because I thought trolls, even childish trolls, occasionally must be confronted. Perhaps I was mistaken about that.

[ Parent ]
Wow, you said that magic 'troll' word. (3.66 / 6) (#88)
by RobotSlave on Tue Aug 19, 2003 at 10:42:09 PM EST

Do you think if you repeat the word "troll" often enough, my valid criticisms of your article will just vaporize? Do you realize that your approach is a textbook example of what they call the "name-calling fallacy?"

Please decide which sputtering excuse you're going to stick with. Are you trying to tell us the history of Sun, and failing miserably by ignoring the first half of that history, or are you trying to tell us about the rise and fall of Sun's stock during the dot-com bubble, without any reference at all to speculation in tech shares?

Either way, you've done a lousy job.

[ Parent ]

YAAT. (4.66 / 3) (#118)
by ckaminski on Wed Aug 20, 2003 at 01:01:31 PM EST

He specifically mentioned, not once, but twice, the significance that the tech bubble stock speculation had in the significant rise and fall in the market capitalization of Sun Microsystems.  If you had read the article, you'd know that.

If you're so upset with his treatment of the first half of Sun's history (which is irrelevant to this article), you can feel free to write your own damn article to educate us all.

Have a nice day.

[ Parent ]

Oooh, an acronym! (none / 0) (#153)
by RobotSlave on Fri Aug 22, 2003 at 05:46:40 AM EST

Wow, I'm definitely defeated now. There's no arguing against acronyms, you know.

Now tell us, ckaminski, do you think slashcart's abysmally bad article is a botched history of Sun, or a botched explanation of the rise and fall of Sun's share price in the dot-com bubble?

Please explain the following quote in your response:

"Thus, Sun's fantastic ascent was largely due to three factors: first, the abandonment of the Unix market by Sun's competitors; second, the technological failures of the Intel/Microsoft juggernaut; and third, the Renaissance of Unix"

[ Parent ]

Agree, kinda (4.50 / 8) (#56)
by johnny on Tue Aug 19, 2003 at 12:41:02 PM EST

I agree with the gist of your comment, even if the tone is unnecessarily snide.

I started working at Sun in January, 1986 and left nearly nine years later, just about the time that, according to this article, Sun's "remarkable ascent" was supposed to be starting. During the time that I was there Sun went from about 1,000 to 15,000 employees, and revenues went from something like $50million/year to $5 b-b-b-billion/year. If that's not a remarkable ascent, I dont' know what is.

The mighty giant DEC was humbled by Sun, and IBM itself came damn near to extinction. It would be stretching things to say that Sun nearly killed IBM, but Sun was a big part of the onslaught that nearly killed IBM.

Sun's other UNIX rivals did not simply walk away from UNIX because they read a few anyalyst's reports. That indeed is fatuous nonsense. They may, however, have decided not to duke it out with Sun, because Sun had a pretty formidable reputation as a street fighter and giant killer. Sun's ascent in the mid-nineties was made possible by 12 or so years of establishing itself as the king of the UNIX hill. They didn't just fall off the apple cart and claim a market by default.

(Now, on the other hand, I would be interested to know how DELL took over their market. That seems to me to be one company that went from nowhere to world dominatin overnight. But I don't know anything about them, so that may be a mistaken impression.)

Hindsight is always 20-20, by the way. So now pundits are saying that Sun's strategy was always to make money on iron, and that they only did software because nobody else would. This impression is absolutely false, at least as it pertains to Sun's strategy in the early 90's. I can tell you first hand how wrenching it was when Sun spllit itself into SMCC (Sun Microsystems Computer Corporation) and SunSoft (and a bunch of other smaller companies).

The reason that Sun went to all the trouble of surgically cutting itself into pieces was because the margins were so much higher in software than they were in hardware, and Sun, which came into being as a mini-me to Digital, was trying to reinvent itself as a new MicroSoft. The strategy was to make Solaris the enterprise standard on ALL hardware, and Sun was trying hard to get HP, IBM, Compaq, and others, to make the switch. There was a lot of speculation that Sun would sell off its hardware business in order to convince the other vendors that it was not self-dealing. Obviously the strategy failed, but it wasn't for lack of trying.

It's been about a decade since I left so my analyses of recent doings are no more valid than anybody else's, but I do get the impression that Sun management killed the goose that laid the golden egg. That goose was its employees and the corporate culture of risk-taking and fun.

Sun became increasingly sclerotic and mercenary between 86 and 94. When I started there, McNeally would rally the troops with talk about how we were going to reinvent computing and turn the industry on its ear. By the time I left he had become a broken record who talked about "increasing shareholder value" and nothing else that I can remember. Employees were demotivated and fearful of layoffs, which were ruthlessly carried out at random intervals. It had become just another company.

yr frn,
jrs
Get your free download of prizewinning novels Acts of the Apostles and Cheap Complex Devices.
[ Parent ]

not sure I agree... (4.16 / 6) (#64)
by slashcart on Tue Aug 19, 2003 at 01:13:10 PM EST

I'm not sure I agree with everything you said.


I started working at Sun in January, 1986 and left nearly nine years later, ... During the time that I was there Sun went from about 1,000 to 15,000 employees, and revenues went from something like $50million/year to $5 b-b-b-billion/year. If that's not a remarkable ascent, I dont' know what is.

From May '87 to Jan '95 Sun's stock price increased only 30%, underperforming the Nasdaq as a whole. Of course there are different statistics you can look at, but this one doesn't seem to indicate a remarkable ascent during that time period. On the other hand, during the 18 months starting in '98 Sun's value multiplied by 30 times. (This data is available from the historical quote mechanism at finance.yahoo.com).


The mighty giant DEC was humbled by Sun, and IBM itself came damn near to extinction. It would be stretching things to say that Sun nearly killed IBM, but Sun was a big part of the onslaught that nearly killed IBM.

DEC wasn't humbled solely by Sun. DEC never successfully transitioned from selling VAX/VMS boxes to selling Unix boxes. Dec went from VAX/VMS to Ultrix/MIPS to OSF (DEC Unix)/Alpha. That's 3 different platforms accross 15 years.

IBM had some difficulties in the early '90s but was never nearly killed. Even at its nadir it had a market cap of $100b (in 2002 dollars). Sun was not the primary cause IBM's decline. IBM's biggest problem was having 400,000 employees.

Sun's ascent in the mid-nineties was made possible by 12 or so years of establishing itself as the king of the UNIX hill. They didn't just fall off the apple cart and claim a market by default.

In the early '90s the biggest Unix vendor by far was HP. 5 years later Sun had more marketshare than all the other contenders combined.


By the time I left he had become a broken record who talked about "increasing shareholder value" and nothing else that I can remember.
I suppose that probably happens with every company sooner or later, unfortunately.

[ Parent ]
Dear fool: (3.75 / 8) (#74)
by RobotSlave on Tue Aug 19, 2003 at 03:06:15 PM EST

"On the other hand, during the 18 months starting in '98 Sun's value multiplied by 30 times.

There was, like this thing? And it was called a "dot-com bubble?" And, like, it inflated technology shares?

[ Parent ]

curious... (4.33 / 3) (#69)
by slashcart on Tue Aug 19, 2003 at 01:42:33 PM EST

During the time that I was there Sun went from about 1,000 to 15,000 employees, and revenues went from something like $50million/year to $5 b-b-b-billion/year.

Hmm, that's very strange. I'll have to investigate why that didn't affect the market cap of the company very substantially. Who knows, perhaps that's why McNealy suddenly started talking about "maximizing shareholder value."

(As an aside...)


Now, on the other hand, I would be interested to know how DELL took over their market. That seems to me to be one company that went from nowhere to world dominatin overnight.

Dell's success is due to a supply chain innovation. Dell only sells computers build-to-order and the prices of various components are dynamically adjusted to keep component inventory very low. That's partly why Dell's website throws random deals at you like "Free CD burner if you buy now!" It means there's a glut of CD burners.

Because computer components decline rapidly in price, they lose something like 5% of their value every month. Dell has a 3-day inventory and Compaq has a 1-month inventory; thus Compaq loses more money from the rapid depreciation of inventory. This allows Dell to sell computers for like $50 cheaper. That might not seem like much but it makes a big difference in a low-margin business.

[ Parent ]

Oracle's Linux strategy (4.33 / 3) (#70)
by ckm on Tue Aug 19, 2003 at 02:33:43 PM EST

Oracle is rapidly killing the need for Sun. The Oracle 9i RAC product is what they are pushing for HA computing, and it's a grid computing system using commodity hardware and Linux.

See Oracle 9i RAC

Here's what Larry said about Linux:

"Linux is a spectacular new answer to some of the issues that are facing enterprise IT managers today--especially the challenge to deliver more performance while saving money in this tough economic environment. After all, Linux has transformed our own business and we're promoting it because we think it's the cheapest, fastest, and most reliable system around."
--
Oracle CEO Larry Ellison
(from http://www.oracle.com/technologies/

Bottom line is that the need for large hardware is rapidly going away, and with it Sun's market for the E10k and E15k servers.

I work for an Open Source strategy company (Olliance Group) and we see people moving to Linux and Linux clusters every day.

Chris.

[ Parent ]

I can't believe I gave you a 5. (none / 0) (#102)
by porkchop_d_clown on Wed Aug 20, 2003 at 09:23:10 AM EST

My finger must have slipped.

:-P

You're right though, DEC is the classic example of superior tech failing in the market place. At given clock speed, Alpha is still (7 years later) the fastest CPU on the planet.


--
You can't raise my prices. You can't build more power plants. You can't build more power lines. Why are my lights out!?!


[ Parent ]
Too Software Focused (4.00 / 2) (#37)
by Silverfish on Tue Aug 19, 2003 at 09:10:21 AM EST

Sun, and all the other major UNIX vendors are primarily hardware companies, who made their OS to sell their hardware.  Some UNIX vendors now sell hardware with Linux available as an option, for example.

The real problem is not related to Operating Systems, it is related to hardware cost and performance.

Solaris is a good UNIX, but Sun never fully commited to selling the PC version because they were too interested in selling their hardware.

yup (none / 0) (#50)
by asad on Tue Aug 19, 2003 at 12:17:38 PM EST

in fact they decided to abondon it but they realized how many customers they had.  I still see Solaris on x86 out there, a lot of people like the idea of keeping a common platform  across all their hardware.

[ Parent ]
Sun's rise and fall? (2.66 / 6) (#39)
by Akshay on Tue Aug 19, 2003 at 09:15:58 AM EST

I see it daily, except when it is rainy, of course.

Sunset strategy (4.60 / 5) (#43)
by rleyton on Tue Aug 19, 2003 at 10:37:33 AM EST

In the finance industry  'Sunset strategy' is a term often referred to. It's a good example of how a major industry is moving away from Sun.

The finance industry is perhaps the biggest single market for Sun kit historically. If you've ever walked around the datacentre of one of the major merchant banks, it used to be full of Sun kit (up to starfires, several in one DC in many cases). It still is to a certain extent.

Ok, so the dot.com boom took Sun's attention elsewhere, but much of their base revenue came from banks (and still does). In addition to buying kit, there's the not-insubstantial revenue from support contracts. As a finance admin, I do not want to wait 24 hours for a new disk or motherboard. I want somebody on site to replace it within a few minutes. So, banks pay a lot of money to Sun for such kit and staff to be on-site. It's much cheaper than lost trades.

Sun kit is more expensive than commodity systems, so CTO/CIO level decision makers are looking at their expenditure and planning accordingly. Getting rid of Sun kit, moving to commodity (read Intel/AMD PC systems) and cheaper maintenance (same kit as PC's), stocking costs and staff costs go down.

Linux is being adopted by many of the banks as a medium term replacement for Sun/Solaris platform. Oracle has recently moved their internal systems off of Sun, a big shift (Oracle and Sun used to be buddies of the first order). Financial applications (Murex etc.) are supporting Linux now, Reuters too. Not much that's not been ported that banks use and rely on.

I used to be - and still am to an extent - a big fan of Solaris systems as solid stable solutions for back-end mission critical systems. However, Linux systems on cheaper - commodity - hardware (often just as reliable, if not more so), designed for failure, make for a compelling alternative.

Bottom line, you're right; Sun are facing a huge challenge. Watch what happens in Finance as an indicator. More banks are talking of walking away from Sun. Sun's revenue will get seriously eroded over time as a direct result, until they find other revenue streams. And this sector is key to Sun.

--
Ooooooooooooooh! What does this button do!? - DeeDee, Dexters Lab.
My Website

hmm.. (none / 0) (#45)
by slashcart on Tue Aug 19, 2003 at 11:07:21 AM EST

I regret that I'm not that terribly embedded in the financial industry. I'm curious to know what retired high-end servers are now being replaced with in that industry: E15k's, HP Superdomes, IBM p690s (or z990s), or whatever else.

[ Parent ]

Smaller boxes... (none / 0) (#67)
by rleyton on Tue Aug 19, 2003 at 01:28:48 PM EST

I wouldn't say that they're planning on replacing boxes in the E15k range just yet, but it's more the low-end to middle-range systems (1-8 cpu systems) that are being, or likely to be dropped, in favour of smaller, faster cpu, intel/amd systems. Commodity, high performance CPU's leave Sparc standing, as you say. Moore's law also eats somewhat into CPU capacity. You can argue that if cpu 'performance' (well, transistor density) doubles every 18 months, you could halve the number of CPU's and get similar performance in many cases.

Don't get me wrong, "sunset" doesn't necessarily imply a complete abandonment of Sun (remember, even mainframes are still used by many institutions), it's just that it's not going to feature as the predominant platform of choice as previously.

Blade solutions are also a serious and growing alternative to "big iron" large SMP boxes. Properly designed application infrastructures that can utilise smaller systems make a lot of sense. That in itself is a big threat to Sun too, who make money on big iron, and whose blade offering is only now making an appearance.

--
Ooooooooooooooh! What does this button do!? - DeeDee, Dexters Lab.
My Website
[ Parent ]

the other way (5.00 / 1) (#49)
by asad on Tue Aug 19, 2003 at 12:16:07 PM EST

"However, Linux systems on cheaper - commodity - hardware (often just as reliable, if not more so), designed for failure, make for a compelling alternative. "

What Intel/AMD combination do you know that allows for hotswapping of CPUs ?  Sun has had howswapping for a very long time and it's something that the x86 platform still hasn't caught up with.  True some people are migrating away from Sun because of the price/performance ratio but no one is going to Linux for reliability over Sun.

[ Parent ]

Not a problem (none / 1) (#60)
by vetgirig on Tue Aug 19, 2003 at 12:51:42 PM EST

Often one replace one Sun server with several Linux servers. Hardware failure in one Linux server means you have to replace the failing part in that server meanwhile the other servers continue to serve your needs.

With several machines doing the same task you do not need for all of them to be up at the same time.

Another solution is using IBM iron that has hotswap capability.

[ Parent ]

several boxes != SMP machine (none / 0) (#87)
by asad on Tue Aug 19, 2003 at 10:05:32 PM EST

So in effect you are saying you would replace a Sun box with a cluster of linux boxes at which point are you really saving money ? Now why would I move to IBM from Sun ? So that I can use linux ? Linux over Solaris ? That's a whole other discussion but I'd still stick with solaris, it's a much more mature OS than linux.

[ Parent ]
I fail to see that as a good thing. (none / 0) (#59)
by joto on Tue Aug 19, 2003 at 12:51:14 PM EST

However, Linux systems on cheaper - commodity - hardware (often just as reliable, if not more so), designed for failure, make for a compelling alternative.

"Designed for failure"? Really? Who would do such a thing? And who would use something that is designed to fail?

[ Parent ]

If I had to make a guess (none / 0) (#62)
by fuzzysteve on Tue Aug 19, 2003 at 01:02:17 PM EST

I would assume that what was meant there was: 'designed so failure was not an issue'

That's the meaning I took from it.

A high availability cluster constructed from commodity parts, is often cheaper than a fully redundant system. Simpler as well.
Bureaucracy destroys initiative. There is little that bureaucrats hate more than innovation, especially innovation that produces better results than the old routines. Improvements always make those at the top of the heap look inept. Who enjoys appearing i
[ Parent ]

Recovery Oriented Computing (5.00 / 1) (#66)
by rleyton on Tue Aug 19, 2003 at 01:22:20 PM EST

You might want to check out the work of David Patterson on Recovery Oriented Computing. It might sound cheesy and a little obvious, but if you design for failure, ie. take it as read that failure is inevitable, then you can build more reliable systems as a whole.

--
Ooooooooooooooh! What does this button do!? - DeeDee, Dexters Lab.
My Website
[ Parent ]
Hmmm,,,, (none / 0) (#82)
by joto on Tue Aug 19, 2003 at 07:41:16 PM EST

Ahh, well. I would say that "designing for failure" fails to communicate what's really intended then.

In any case, building more reliable systems is what we all should do, and expecting failure should be part of any risk-analysis, so I fail to see anything new here. (But some of the projects on the page you referred to me looked quite neat, so I don't really care what you call it).

[ Parent ]

So what ever happened to EPIC? (3.00 / 1) (#48)
by Merc on Tue Aug 19, 2003 at 12:04:19 PM EST

Are modern Intel processors like EPIC was supposed to be, 5 years ago? Are other companies going that route?

Although things may not have worked out for Intel in the mid 90s, the concept still seems good. Compilers should be doing as much work as possible, so that all the "guessing" that modern processors do is more accurate.

Does anybody know what kind of architectures are going to be common in desktop computers in 2 or 3 years?

The other thing that I wonder about is what this mistake by MS and Intel will mean in the long run for MS. Had they come out with a good OS running on an EPIC processor, Unix might well have died, but they didn't, and it didn't. Instead, Linux appeared. Since nobody ever really fully moved away from Unix, maybe they never will now. Maybe MS is doomed because they missed their chance to get rid of Unix. If they had an OS that was fundamentally better than a Unix-style OS, maybe it would just be a delay. The way it looks now, Unix is still the best architecture, and you can now get it (in the form of Linux) for free.



EPIC's fate (3.00 / 1) (#54)
by slashcart on Tue Aug 19, 2003 at 12:31:55 PM EST

Although things may not have worked out for Intel in the mid 90s, the concept still seems good. Compilers should be doing as much work as possible, so that all the "guessing" that modern processors do is more accurate...
The concept seemed excellent, which is why EPIC had such an impact even before it was released. The difficulty is, with branch-intensive code it simply isn't possible to predict the code path taken at compile time. This (from what I understand) is why Itaniums perform so well on floating point code (loop-intensive) and not as well on commercial workloads (branch intensive). Intel's trying to mitigate this by using profile-driven compiler optimization. Using this technique, a program is compiled and run, then profile data is gathered about which branches are taken etc, then the program is compiled again using the profile data as a guide.


Does anybody know what kind of architectures are going to be common in desktop computers in 2 or 3 years?

Intel will soon be releasing a version of Itanium 2 for workstations, small servers, and desktops. This version (called "Deerfield") will have a 1MB cache with a die size and manufacturing cost similar to Xeons. Apparently the price will be approximately the same as the 3GHz 1MB Xeons. Since Itanium2 has much better fp performance than any x86 or Sparc box it's quite possible that deerfield and derivatives will take a large portion of the technical/scientific workstation market over time. More than one person has speculated that Intel plans to eventually transfer the entire desktop market to chips with the Itanium ISA. That would probably take a number of years. I'm not certain what AMD would do at that point.

[ Parent ]

Re: EPIC's fate (none / 0) (#154)
by Amorsen on Sun Aug 24, 2003 at 11:38:59 AM EST

More than one person has speculated that Intel plans to eventually transfer the entire desktop market to chips with the Itanium ISA. That would probably take a number of years. I'm not certain what AMD would do at that point.

Die of laughter, perhaps? The Itanium ISA is so large and complex as well as dependent on fancy hardware to keep a decent performance, that it will be many years before an implementation for the desktop market is viable.

Intel lacks a 64-bit chip for workstations and desktops. Right now the pain is only moderate, but in two years it will be a serious problem. I predict that a solution will be found in less than two years, and I predict that it will not have IA-64 as its primary instruction set.

[ Parent ]

not sure you're correct about this (none / 0) (#155)
by slashcart on Sun Aug 24, 2003 at 05:16:31 PM EST

Die of laughter, perhaps? The Itanium ISA is so large and complex as well as dependent on fancy hardware to keep a decent performance, that it will be many years before an implementation for the desktop market is viable.

Intel is releasing its "deerfield" workstation Itanium2 processor in 2003, according to Intel's roadmap. This deerfield processor is for workstations and small servers, and will cost about the same as the high-end Xeon.

The ISA is not dependent on fancy hardware for decent performance. Instruction scheduling etc is done in the compiler, therefore the Itanium isa is dependent on the compiler for adequate performance.

Without caches the deerfield die size is actually smaller than a p4's and the chip is substantially less complicated. This is partly because the i2 doesn't require logic for OOO, or for translating baroque x86 instructions into simpler micro-ops.

[ Parent ]

Sure Itanium-2 is smaller without caches... (none / 0) (#156)
by Amorsen on Sun Aug 24, 2003 at 07:01:34 PM EST

It is also entirely dependent on those caches for performance. The P4 needs memory bandwidth but can work around high latency. Itanium-2 needs loads of bandwidth and large, low latency caches.

[ Parent ]
hmmm... (5.00 / 1) (#157)
by slashcart on Sun Aug 24, 2003 at 08:19:28 PM EST

It is also entirely dependent on those caches for performance. The P4 needs memory bandwidth but can work around high latency. Itanium-2 needs loads of bandwidth and large, low latency caches.

I don't have an i2 to run applications and measure performance. But at least on SpecCPU, cache size doesn't seem to have more of an effect on i2 performance than it does on Xeon performance.

The 900 MHz i2's outrun the 3.06GHz Xeons on SpecFP, even though they both have the same cache size (1.5MB). This despite the fact that the i2 is running at only 900MHz, well under 1/3rd the clock speed of the Xeon.

I'd have to guess that, in the long run, the itanium line will at least capture the market for technical/scientific workstations. I2's offer far better floating point performance than x86 or UltraSparc-III, for a smaller die size and lesser manufacturing cost.

[ Parent ]

Intel is trying to head this way... (none / 0) (#57)
by mge on Tue Aug 19, 2003 at 12:43:33 PM EST

Compilers should be doing as much work as possible, so that all the "guessing" that modern processors do is more accurate.

Intel's IA-64 Architecture (a whole new redesign, a departure from x86), is trying to go this way.  I would describe it as a more RISC like architecture which works well with optimizing compilers.

Here's a good overview on the IA-64: http://www.arstechnica.com/cpu/1q99/ia-64-preview-1.html which might explain where Intel is headed.

[ Parent ]

IA-64 == EPIC (none / 0) (#61)
by achtanelion on Tue Aug 19, 2003 at 12:59:58 PM EST

enntee

[ Parent ]
links (none / 0) (#58)
by asad on Tue Aug 19, 2003 at 12:45:05 PM EST

some links would have been nice for the story.
Here a few from the Register

Suns upcoming processors

Faster chips for the V480


Here are links... (none / 0) (#132)
by slashcart on Wed Aug 20, 2003 at 07:59:55 PM EST

I provided some links in this post.

[ Parent ]
cool (none / 0) (#133)
by asad on Wed Aug 20, 2003 at 11:02:07 PM EST

why didn't you include them in your main article ?

[ Parent ]
Good timing (5.00 / 1) (#71)
by onyxruby on Tue Aug 19, 2003 at 02:42:41 PM EST

Nicely written, good timing, as I'm presently scheduled to start school for SUN / Solaris here in September. I've already been through the MS / Cisco route and have several years IT experience. I have been wondering for a while now if SUN is going to be around in several years. I want to gain a solid Unix skill base, and certainly it's hard to deny Solaris' impact historicaly, but than you could argue the same thing about DEC / Alpha, and that architecture got put out to pasture by Intel a little while back.

Does anyone have any solid opinions on SUN / Solaris as being a worthwhile educational route? My frame of thinking is that Solaris would give me a solid Unix base that could be used to pick up other *nix's fairly easy. I'm also going to pickup a piece of their kit off Ebay for hack purposes whilst going to school. I have a rack that will hold SUN equipment, any suggestions on what to put in it? I would have a budget of about a grand, and have had trouble googling good information about a learning - not production platform.

One other thought, does anyone else think SUN's fascination with trying to convince people to go back to terminal computing backfired?

The moon is covered with the results of astronomical odds.

Learn linux (none / 0) (#73)
by Julian Morrison on Tue Aug 19, 2003 at 03:00:57 PM EST

It's far cheaper, and less likely to do a DEC. And once you know linux, shouldn't be hard at all to transition to solaris or any other commercial unix.

[ Parent ]
Try convincing... (none / 0) (#84)
by AngelKnight on Tue Aug 19, 2003 at 08:21:13 PM EST

an American Human Resources drone that the skills can translate over *wink*

[ Parent ]
Yes and No. (none / 0) (#147)
by Lin Dze on Thu Aug 21, 2003 at 09:29:41 PM EST

Learning linux is probably a bit cheaper in the short term compared to buying a Sun Ultra or 2.

But when you go to an interview would you rather say:
"I have Admin experience running a RedHat server and multiple clients"
or
"I currently admin 2 Sun Ultras and a group of Sparcstations."

The Ultras might cost a bit more off ebay, but Solaris IA32 can be set up for the same trivial cost as the Linux stations and has a WHOLE lot more credibility with management. Fact is theres still a lot of bias for "big iron" vendors over linux.

-Lin Dze
"Facts don't cease to exist because they are ignored." Aldous Huxley
[ Parent ]

Are you talking about vocational qualifications? (none / 0) (#78)
by it certainly is on Tue Aug 19, 2003 at 05:55:21 PM EST

Just wondering.

kur0shin.org -- it certainly is

Godwin's law [...] is impossible to violate except with an infinitely long thread that doesn't mention nazis.
[ Parent ]

Yes and no (none / 0) (#85)
by onyxruby on Tue Aug 19, 2003 at 08:33:03 PM EST

Hard question to answer. I'm about three quarters of the way to a bachelors, but the credits are in all the wrong places. I do want to end up getting it someday, but at the moment I'm more looking to get knowledge and skills that I'll need to develop my career in IT. The basic problem I've seen with four year colleges so far is that they don't teach you anything practical about computers. With the job market as bad as it is, I can't afford to wait a few years to pick it up in some Uni.

You can get a degree that centers around programming, which I find incredibly boring, or electrical engineering, but I have no interest in being an EE. What I want to do is work in IT NOC's. I have yet to hear of any college that has an actual bachelors in networking outside of the U of M. Unfortunately the last time I checked they required more math for their BIN program than they did for their engineering programs. Once I can get enough to start working with *nix, I can continue my everlasting quest to finish college:)

The moon is covered with the results of astronomical odds.
[ Parent ]

Go for it! (5.00 / 4) (#91)
by Global-Lightning on Wed Aug 20, 2003 at 12:24:11 AM EST

Great plan.  Getting Sun hardware is great for learning:
1.  It will expose you to a different hardware architecture.  A 64-bit platform that supports hotswapping... You'll also see the design limitations and capabilities compared to commodity x86 stuff.
2.  In addition to Solaris, you can load Ultra Linux, NetBSD, and OpenBSD. Learn the difference between System V, BSD, and linux flavors.  Next to x86, Sun hardware is the most supported by the Open Source movement.
3.  The "cool" factor.  For the cost of an old Pentium, you can impress your geek friends with a genuine UNIX workstation at home :)

For the money I would get an Ultra 60 or a Blade 100.  The Ultra seem to be the sweet spot for price vs performance. For less than $1K you can pick up a dual processor Ultra 60 with decent memory and SCSI disks.  This box will give you decent performance for web surfing, openoffice apps, and moderate compiling.  Ultra 30, 10, 5, and 2's can be bought for a couple hundred bucks, not as capable but still nice. The old Sparkstation "pizza boxes" are incredibly cheap but you'll outgrow them quickly.  The Blades have faster CPU's but IMHO aren't that much better than Ultras.

Things to watch out out for:
1.  256K is the absolute minimum for memory.  The more the better!
2.  Monitor connections - most older sun workstations use 13W3 connectors instead of the more common 15HD used on PCs.  There are 13W3/15HD converters.  Sun also sells monitors that take both standards and lets you switch between your Sun and PC.

At work I run an Ultra 60 next to my Dell, feeding into a Sun 21" monitor.  I love it, and will buy one myself soon.

[ Parent ]

Tee hee (3.50 / 2) (#97)
by synaesthesia on Wed Aug 20, 2003 at 07:49:15 AM EST

Things to watch out out for:
1.  256K is the absolute minimum for memory.

But 640K ought to be enough for anybody!


Sausages or cheese?
[ Parent ]

Sun is doomed (2.25 / 4) (#75)
by riceowlguy on Tue Aug 19, 2003 at 03:12:35 PM EST

The main reason is that all software written for Solaris (or any other Unix box other than Linux, it seems) more complex than xemacs on Solaris is an agonizing pain to use and considerably more difficult to administer than its Windows counterpart. Seriously. I just was trying to help somebody this morning set up spam filtering on her Netscape 7.0 for Solaris. Clicking on "New" in the filters dialog box caused the whole application to freeze up. That is simply apalling.

An e.g. from my particular locale's area of expertise: all the best programmers at the enormously wealthy Landmark (i.e. Halliburton) and Schlumberger haven't been able to make their respective seismic interpretation packages anything other than massive, thrashing peices of shit in all the years they've been at it, whereas SMT, a relatively small newcomer to the industry, has a product that totally destroys the others, quality-wise. So much so, in fact, that Landmark has tried to buy out SMT several times. The reasons are obvious: Windows programmers only have to test their stuff on one OS (or, since the introduction of 2k and XP, three variations of the same OS). That, and X-Windows bleauxs geauxts.

"That meant spending the night in the living room with Frank watching over me like some kind of Lovecraftian soul-stealing nightmare creature-Azag-Frank

More complex than XEmacs (none / 0) (#92)
by mysta on Wed Aug 20, 2003 at 01:24:30 AM EST

The main reason is that all software written for Solaris ... more complex than xemacs on Solaris is an agonizing pain to use and considerably more difficult to administer than its Windows counterpart.
More complex than XEmacs?

That's setting the bar pretty high. XEmacs isn't exactly the simplest piece of software in the codesphere.
---
Are we not drawn onward, we few, drawn onward to new era?
[ Parent ]

Well, you know, atom bomb simulators (none / 0) (#100)
by porkchop_d_clown on Wed Aug 20, 2003 at 09:17:11 AM EST

stuff like that.

All though, I have to admit there's probably slightly more lines of code in Office than EMACS.


--
You can't raise my prices. You can't build more power plants. You can't build more power lines. Why are my lights out!?!


[ Parent ]
Yeah, well, that's another problem [nt] (none / 0) (#120)
by riceowlguy on Wed Aug 20, 2003 at 01:38:55 PM EST


"That meant spending the night in the living room with Frank watching over me like some kind of Lovecraftian soul-stealing nightmare creature-Azag-Frank[ Parent ]

Said the MSCE. (none / 0) (#101)
by porkchop_d_clown on Wed Aug 20, 2003 at 09:18:29 AM EST

My windows box requires more time and energy to keep virus free each month than I spend in a year maintaining my Linux and Mac boxes. My previous experience with Solaris was the same.


--
You can't raise my prices. You can't build more power plants. You can't build more power lines. Why are my lights out!?!


[ Parent ]
Logical fallacy (none / 0) (#122)
by riceowlguy on Wed Aug 20, 2003 at 02:06:03 PM EST

People are always assuming that there are no Unix or Mac viruses because of the inherent security features of those operating systems. In reality, there are no Unix or Mac viruses because nobody can be bothered to write them. Seriously. If I write a virus for Windows, I've covered 99% of all my potential targets for an average of 33% of the work.

"That meant spending the night in the living room with Frank watching over me like some kind of Lovecraftian soul-stealing nightmare creature-Azag-Frank[ Parent ]

Partly true. (none / 0) (#143)
by porkchop_d_clown on Thu Aug 21, 2003 at 08:24:39 AM EST

"Back in the day" I saw viruses on the Mac, Amiga, and so on. And it is irrefutable that worms regularly attack linux boxes.

But MS in their rush to add features regularly ignores basic security design. Why were these ports even open on my machines? Not because of any feature I wanted - or was even aware of. Meanwhile, OS X starts with everything turned off and you enable those features you want - you don't have to hunt down and turn off the features you don't want.


--
You can't raise my prices. You can't build more power plants. You can't build more power lines. Why are my lights out!?!


[ Parent ]
Windows development (none / 0) (#108)
by bored on Wed Aug 20, 2003 at 11:40:15 AM EST

The reasons are obvious: Windows programmers only have to test their stuff on one OS (or, since the introduction of 2k and XP, three variations of the same OS).

It doesn't hurt either that the higher level API's on windows are 'standard' and included in the basic OS package. Creating a GUI component for windows is 100x easier than doing it in java or C++ for X/windowmanagerofchoice. Nor does it hurt that the good development tools are 'better' for windows. One look at a bounds checker (windows only) almost proves its better than purify (cross platform). The concept of an IDE is inheritly more efficient because you can write, debug, test, manage everything from a single point. Sure unix can do that too, but only after you have spent days configuring emacs/vi/visual slick edit/etc to be usable. Delphi is awesome for GUI development. Windows development may be slightly harder to learn but once you get it down, what you can do is truely amazing. Plus there is the programmer inertia. If the majority of the programs written are for windows, then its more likely that there are more 'good' programmers working on windows projects. Its not like the best programmers are automatically unix guys. It could probably be argued that since the unix programming model (text based io streams) attracts the programmers who can't deal with the more complicated windows models (message passing) and are therefore less skilled.



[ Parent ]
Developpers. (none / 0) (#127)
by Tezcatlipoca on Wed Aug 20, 2003 at 06:21:30 PM EST

The windows developpers I have to deal with do not understand the limitations and the stress on the resources they are using.

The UNIX programmers do.

Don't tell me who are better programmers, Solaris vs Windows, Apache vs IIS, Oracle vs MSSQL, pine vs Outlook has the right answer for me.

Might is right
Freedom? Which freedom?
[ Parent ]

Obviously your not admining serious stuff (none / 0) (#119)
by Alhazred on Wed Aug 20, 2003 at 01:12:30 PM EST

Windows requires something on the order of 10x the admin time of an equivalent *nix box. Between that and abysmal reliability, plus the fact that you can't scale NT/2k/XP for kaka and you have a real winner.

Also if you aren't a developer, don't try to comment on the ease of development or reasons for quality/lack of quality of software on a particular platform. There is no technical reason why *nix applications would be inferior. If a particular vendor is lame, then that vendor is lame. You also don't seem to understand that different markets demand different feature sets. Nobody that runs Oracle for instance gives a rat's patootey that it takes a DBA to install it, thus 'ease of installation' is not a very high priority for them, but you can DAMN well bet that while MS SQL server is easy to install it is TOTALLY HAPLESS at addressing the kinds of needs high-end Oracle users have.

It certainly is possible that in 20 years MS will have wedged Win 2020 into data centers, but I sure ain't holding my breathe. *nix systems get the job done.
That is not dead which may eternal lie And with strange aeons death itself may die.
[ Parent ]

Oh please. (none / 0) (#126)
by Tezcatlipoca on Wed Aug 20, 2003 at 06:16:45 PM EST

In the immortal words of Johnny McEnroe "you cannot be serious man!".

From all the examples that you may have used to bash Sun and Solaris you use the most inconsequential of all examples, a fscking web browser. An application. Not even the OS. I will not criticize the wisdom of filtering spam at the workstation level, but frankly companies using Solaris are not worried about a web browser (PS: I have been using Mozilla since version 0.7 with no problems whatsoever. The mail filtering is done at a higer level since we take very seriously email safety).

Now, let me elaborate a bit. I have a few DNS servers that keep a big corporate network running. Uptimes? Months. There is no recorded crash for these machines and I have been here already 3 years. Ditto for corporate DHCP servers. This is not the typical Linuxy "l00k at my 33lit3 up7im3", this is a fact that we have not lost money due to these Sun machines not failing.

Some other servers have proven more troublesome because we have stretched them to the maximum. Guess what, Sun has been with us holding meetings, releasing patches for us and explaining to very fine technical detail what they are doing. Short of showing us what they did in the source code, they have done all what is to be expected from a company at this level.

Hardware? In the very rare ocurrences of a hardware problem the on site Sun guys sort it out for you in less than 24 hours.

When I had the misfortune of not having on site Sun personnel I always got spare parts in less than 48 hours. That is specially valuable if you are in an office in the outskirts of Hanoi and a Charisma set up (you know Charisma I guess) has failed due to a catasthropic problem (water, flooding, etc).

Now, going back to the oil industry, most companies and geoscientists must be pretty ignorant since without excepcion in my personal experience they love SLB stuff (specially Charisma), tolerate Landmark and don't know about SMT. After so many years in business and remaining so small I think one can assume that SMT's offerings are not that revolutionary, and knowing the differences of stability, reliability, scalability and performance between Wintel and Solaris, it is not surprising why they remain small while Landmark and SLB can keep swallowing small players so easily.

Might is right
Freedom? Which freedom?
[ Parent ]

Brilliant Logic (none / 0) (#138)
by mickey mouse on Thu Aug 21, 2003 at 04:03:28 AM EST

My friend, the most complex thing you run on computer is obviously Xemacs :-). The stability of a *nix system is infintely more than that of any M$ platform that I have seen. I use a Sun workstation that I have not rebooted in the last six months. Rebooted at that time because of a hardware failure. I have not logged out of my system for the last 4 months. I am sure there are people out there who can provide statistics for a M$ system. I guess it is easy to write software for M$, if you dont care whether the system stays up past 5 mins . I am obviously exaggerating here but I think you get the point. The reason it might be more difficult to get applications running on Sun or other commercial UNIX boxes is probably because of non availibity of GUI libraries ( although even that is changing very fast).
I work for a chip design company and I am yet to hear about any chip design company using M$ for their work. I have run applications that consume 12-15 GB of memory. I think you can only dream of running something that complex on a M$ machine.

[ Parent ]
That's funny... (5.00 / 4) (#89)
by MKalus on Tue Aug 19, 2003 at 11:54:45 PM EST

I doubt very much that Unix in general will go away. They said the same thing about Mainframes 20 years ago and yet they are still there.

Granted, there is only one company left who provides an OS and only a handful who provide Hardware, but in essence it still exists.

Will the Sun go down?

I dont' think so. What everybody here seems to forget is that SUN is not only in the Tech industries. Banks, Insurance Companies, etc. are ALL using Sun Equipment.

Yes, speed is "of the essence" but it isn't the end of all, sure IBM and HP are both back in the game but the reality is that IBM these days will sell you ANYTHING that you want, even SUN and HP equipment.

HP meanwhile is pretty busy trying to figure out where the "New HP" is going and a lot of people I know who worked for Compaq or HP before are gone.And most people I know who worked with HP products for 20+ years (the "old folks") tell me constantly how much the new stuff sucks, from printers, to scanners, to servers. HP is getting cheap even in the highend stuff.

The only thing I see in masses in the datacenters these days are HP and Sun (not necessarily in that order though). Sun might not make the billions they made during the dot com bubble but they will still be around 10 years from now, I can't really see them go down the drain.

For big companies it is support that counts, availability and experience of their own staff, that's one of the reason why mainframes are still around and one reason why people still by EMC equipment even though the Hitachi solutions are technology wise far superior.

Windows2000, XP and 2003? Well, let's just put it this way, today at work they had to take large parts of our Infrastructure down because of "urgent system upgrades to the Windows Environment", so I doubt we'll see anything REALLY  important on any Windows type anytime soon. A couple of months ago the Slammer worm made it into our Network and did some rather nasty things as well, again Windows. So no, I don't see Microsoft killing of the Unix Equipment we have, which is largely HP & Sun, some (legacy) AIX and DEC.
-- Michael

I agree (3.80 / 5) (#94)
by slashcart on Wed Aug 20, 2003 at 01:42:44 AM EST

I doubt very much that Unix in general will go away. They said the same thing about Mainframes 20 years ago and yet they are still there.

I agree. If commercial Unix did ever die it would likely be because it was replaced by Linux.

Mainframes are not a completely analagous case, however. There's a large body of non-portable software written specifically for the peculiar architecture of S/390. This guarantees continued customers through "lock in." With Sun, however, most of its boxes are running Oracle. Applications interact with the Sun servers at a very high level. It's not so difficult to replace Sun's servers running Oracle with somebody else's servers running Oracle.

Windows2000, XP and 2003? Well, let's just put it this way, today at work they had to take large parts of our Infrastructure down because of "urgent system upgrades to the Windows Environment", so I doubt we'll see anything REALLY important on any Windows type anytime soon.

I'd be surprised if a version of Windows ever penetrated the high end of the enterprise space. The cost savings of using Windows there would be trivial. Windows' benefits (easy GUI administration) are unimportant in that space. There's no reason to deploy Windows there, not now, and not in the imminent future.

[ Parent ]

Oh, yeah! (3.10 / 10) (#95)
by RobotSlave on Wed Aug 20, 2003 at 05:36:48 AM EST

So now you acknowledge Oracle. Where was Oracle in your writeup? Huh?

Why didn't you make any mention in your "article" of the clear and substantial part that Oracle played in the dot-com bubble, and, not coincidentally, in the Sun market-cap "rise and fall?"

Well, this question really answers itself, doesn't it? It's just a tossup between "abysmally poor research" and "flagrant preconceptual bias," so, to be fair, we'll just have to assume it's an even mix of the two.

[ Parent ]

slashcart! (none / 0) (#98)
by synaesthesia on Wed Aug 20, 2003 at 07:54:02 AM EST

What the hell is your problem?

Sausages or cheese?
[ Parent ]
duh (none / 0) (#105)
by Battle Troll on Wed Aug 20, 2003 at 10:39:10 AM EST

Slashcart obviously doesn't like harsh criticism.
--
Skarphedinn was carrying the axe with which he had killed Thrainn Sigfusson and which he called 'Battle Troll.'
Njal's Saga, ca 1280 AD
[ Parent ]
One company left with decent hardware? (none / 0) (#99)
by porkchop_d_clown on Wed Aug 20, 2003 at 09:16:10 AM EST

You mean Apple?<?p>

--
You can't raise my prices. You can't build more power plants. You can't build more power lines. Why are my lights out!?!


[ Parent ]
Support A5+ (5.00 / 1) (#96)
by Wulfius on Wed Aug 20, 2003 at 06:55:21 AM EST

Sun have EXCELLENT support.

I bought a SUN server recently because they were VERY helpful in fixing an old SUN server I had.
I could have gotten an W2K shitbox instead.
The server I got was actually cheaper (Sunfire V100).

I hope they stay around.
Until very recently I had 3 SUN boxes.
I used to point out to the Microsoft bigots
my uptimes on them (Longest was 365 days, interrupted by annual power test).

We are advised by vendors to reboot our Microsoft boxes 1 every 2 weeks.

---
"We must believe in free will, we have no choice."
http://wulfspawprints.blogspot.com/ - Not a journal dammit!

Fujitsu's SPARC (4.00 / 1) (#106)
by Anonymous Hiro on Wed Aug 20, 2003 at 10:58:45 AM EST

Well if you want an arguably better SPARC than Sun SPARC check out Fujitsu PRIMEPOWER:

http://primepower.fujitsu.com/en/index.html

They are getting closer to mainframe-class than Sun is. Their new SPARC chips have hardware instruction retry.
http://www.esj.com/news/article.asp?EditorialsID=291

As for performance check out www.spec.org. They seem to be faster.

AFAIK they're compatible with Solaris.

Dunno about price tho. But Sun is in an unenviable position of being squeezed by Dell etc at the low end, and superior SPARCs and alternatives (HP, IBM) at the mid to near high end.

they actually run solaris (none / 0) (#111)
by asad on Wed Aug 20, 2003 at 12:16:28 PM EST

It's not compatiable it's the exact software, the problem with their boxes used to be that they were basically linking sun boxes together, when Sun had the E10k they had a 128cpu solution but it was 4x as big as a E10k.  Most people who want Sun go to the source not for a copy but it could be that they have gotten better over the last 3 years.

[ Parent ]
Fujitsu's Sparc (none / 0) (#131)
by slashcart on Wed Aug 20, 2003 at 07:52:21 PM EST

An interesting fact is that Fujitsu's SparcIV/64 are fully out-of-order chips running at 1.5 GHz, unlike Sun's, which are in-order and running at 1GHz. This gives Fujitsu's chips substantially better performance on commercial workloads than Sun.

Fujitsu's Sparc boxes run solaris.

[ Parent ]

This just got mentioned on The Inquirer (4.33 / 3) (#107)
by Gedvondur on Wed Aug 20, 2003 at 11:35:56 AM EST

http://www.theinquirer.net/?article=11121

Interesting commentary on this article.

Gedvondur
"Er. You loony bastard, what do you make of this?"--Lance-Constable Detritus

Congrats slashcart! <nt> (1.00 / 1) (#135)
by Vesperto on Wed Aug 20, 2003 at 11:51:36 PM EST



If you disagree post, don't moderate. Alimaniere forf
[
Parent ]
Inaccurate and not particularly well researched (3.85 / 14) (#109)
by irixguy on Wed Aug 20, 2003 at 11:50:37 AM EST

SGI never said they were transitioning to Intel CPUs, and never said they were going to stop development of IRIX. Ever.

Stupid analysts and their parrots presumed all this because SGI launched a (for them) low cost line of Wintel machines.

Next, your comments about the creation of the E10k are bogus.

Cray already had a 64 way Solaris box - the C6400. When SGI bought Cray, SGI found themselves in the poor position of having to support a machine that ran a rivals OS. Part of their deal of buying Cray was to sell that division of Cray to Sun.

The E10k is nothing more than a C6400 with UltraSPARC II CPUs in it. It is not technologically advanced - have a look inside one. They have off-the-shelf 3COM switches inside for their interconnects.

In the same time Sun produced the E10k, SGI produced the Origin 2000. The O2k was far more technically advanced, and significantly faster. It would also scale to 512, and, later, 1024 CPUs - as opposed to 64 in the E10k.

IBM has never 'downplayed' it's power systems - quite the opposite. IBM will sell you whatever you want, but when IBM come in to do a sale that requires mid to high end, you will hear nothing about Wintel, and all about POWER based systems - RS/6000s and AS/400s. This has always been the case through the 90s.

The bizarre yo-yo effect of Sun's stock price was nothing to do with their selling to dot-coms - it was all about analysts overhyping Sun because they had good kit over a wide price range, then slamming Sun because they did not over-sell, then slamming Sun because the dot-com bubble burst.

Sun downplayed their systems to dot-coms, and actively discouraged people from buying kit they did not need.

Post dot-com boom, *all* UNIX vendor's sales plummeted - this is the nature of an economic crash, and not something related soleley to Sun. Due to the above comments about Sun's behaviour during the dot-com years, Sun have weathered this well - their continued poor stock price is, once again, due to analysts bashing them solely because analysts, much like yourself, do not understand a large company with a big R&D budget who solves problems, as opposed to box shifters like Dell.

Sun are still a market leader in the UNIX space. No one will buy HP, because HP's strategy for their UNIX line is highly risky for a company to buy in to.

People who bought IBM kit still do, and some are moving to IBM due to POWER4s performance.

SGI remains the niche player they have always been - they do visualisation and HPC like no-one else.

But companies still buy Sun. Consolidation is the name of the game. Some people think Linux clusters are the way to go. However, consolidating to a large UNIX box is vastly more cost-effective, over the life of the machine - and no-one does a resilient, expandable machine like Sun do.

At the low end, Sun are doing fine. They have machines based on the US IIi and III which will outperform offerings from Dell and IBM - I cite the V100 and friends, and the V480 dominates in the low-end 4 way space, both on price and performance.

Don't be fooled by the muddy thinking of analysts and 'experts' - CPU clock speed is not the be all and end all. Nothing in the class can compare to the V480 for sustained IO throughput - thus it is the fastest solution in that 4 way price bracket.

HP and IBM never 're-committed' to UNIX - they always were committed. HP's strategy is confused, but they have to support all those finance houses using HP-UX, and they have to support Tru64 users for another 16-17 years - due to government contracts. And as I've said, IBM always supported, sold, and pushed their RS/6000s.

Sun is not "one full process generation behind in manufacturing process" - that is twaddle. What you mean to say is that their die size is not as small as Intel - so what? No-one else will have 4 core CPUs next year - pin compatible with US III! They're a generation *ahead* in CPU design.

Another major mistake in this article is your assertion that HP and IBM 'conceded the UNIX market' to Sun. Rubbish. Sun made faster, more cost-effective machines. They still do. They had far better compatability - Solaris is binary compatible across their range - AIX isn't. Sun peripherals will work across their entire range - can't do that with HP-UX boxes.

If, as a CTO, you're faced with buying new cards, new processors, new memory, when moving to the next larger box from HP, or just a new chassis when moving to a larger Sun box, which would you choose?

The US III is not a poor performing chip, by any stretch of the imagination. It has a low clock speed and handles small integer jobs slower than an Intel CPU.

So what?

In the real world, that doesn't matter, unless your running a render farm. Sun's design, like SGI's, keeps the CPUs *busy*, and is optimised for IO, because in the real world, CPUs don't just sit there processing small integer jobs. It's not efficient, and it's not the workload of a business.

In a business environment, a machine with US III CPUs will provide a faster, more cost-effective solution.

And you have completed ignored the support angle. HP support sucks. Any Intel vendors support *sucks* - especially Dell. Sun will get an engineer with parts on site within 2 hours. You can escalate your calls to a senior engineer directly when you log your call, if it is important. And by senior engineer I'm not talking about a Dell MCSE - I'm talking people who wrote the kernel, who designed a SunFire chassis. Helpful people with clue who will do whatever it takes to get you up and running ASAP.

My final point will refer to your 2nd from last paragraph. Sun not only compete, but out-class offerings from Dell in the low-end (see previous comments about the V480). At the high end, Sun rule in the business world. Without question. There are so many features in Solaris and in the F15K chassis that are un-matched by anyone. This means a lower TCO, which means that spending a large amount of Sun kit up-front works out cheaper over it's life than any other offering.

Overall this piece is poorly researched and merely parrots the same inaccurate, poorly thought out analyst comments from the last 8 years.


--

"Tell them we are not Gods, but Sysadmins, which is the next best thing."


you have been drinking too much sun coolaid (none / 0) (#114)
by asad on Wed Aug 20, 2003 at 12:25:16 PM EST

I agree with some of your statments but look at Suns financial statments, their revenu is down.  What does the future hold ?  Better Java ?  The wider newer chips ?  Their grid computing idea has some merit but beside that I haven't seen any major new things come out of Sun and please don't mention N1.  

[ Parent ]
Everyone's revenue is down (none / 0) (#115)
by irixguy on Wed Aug 20, 2003 at 12:36:01 PM EST

And Sun are weathering the storm better than most. As I pointed out, declining revenue for just 1 company would be bad - if everyone's revenue is down, it's to be expected in an economic downturn.

Besides - you've seen nothing new come from Sun? Eh?

Pay attention, lad! Wildfire interconnect allowing you to cluster 4 F15ks or F12ks together is pretty damn big.

SCS 3 is full of goodies that will take advantage of the above.

The new T3 arrays are very sweet pieces of kit - they've fixed most of the shortfalls of the previous range, and price/performance is *very* good.

Solaris 9 (and, with a patch, 8) have a very efficient threading model which provides a significant performance boost when using something like Weblogic or Oracle.

Solaris 10 is coming, and with it a useable container model, as well as processor affinity.

Domain failover and config sharing between SCs that came in the last few SC firmware and RTOS updates is pretty big news. Nothing else out there can do what the SunFire chassis can currently do.

N1 is marketing wank and can safely be ignored for a couple of years :-)

Sun's R&D benefits the enterprise, and this shows in what they're introducing.

I'm not a Sun fan (my username should give that away) - but I am a fan of what works. And right now, Sun are delivering what businesses need. No-one else is.


--

"Tell them we are not Gods, but Sysadmins, which is the next best thing."


[ Parent ]
some of what you said I agree with (none / 0) (#125)
by asad on Wed Aug 20, 2003 at 06:03:36 PM EST

First Sun is a non-player in the storage world, they have tried and tried again but they lost their best NFS people to NetApp and no one really looks at their gear when they think storage now.

The benefits you mention would be useful for big companies that need a big iron box to run oracle or some other dedicated app but that's about it.  
They have lost the speed battle and are now being kicked out of the EDA space.  So in short they could be the big backend box but not the front end box which services the clients directly.
'useable container model' do you mean a java container ? the iPlanet stuff ?

What I'd really like to see from Sun is a big speed improvement.  2 or 2.5 ghz chips in the next year.  

The interconnects would be interesting to see but would that make them a bigger player in the supercomputing space ?  They are currently at 211 in the top 500 supercomputers but without faster chips I don't see them improving that standing a whole lot more.  

At least we both agree on N1.

[ Parent ]

More stuff (5.00 / 1) (#141)
by irixguy on Thu Aug 21, 2003 at 05:12:56 AM EST

First Sun is a non-player in the storage world, they have tried and tried again but they lost their best NFS people to NetApp and no one really looks at their gear when they think storage now.

I would agree with this partly - if I'm speccing a system for clients, I'm going to be talking to EMC first.

However, Sun do some *very* good deals on bundling T3s with their kit. The latest models are actually very good, and sales are looking up.

They also resell Hitachi's high-end gear, which is worth noting only because they're quite good, and they help Sun's income.

The benefits you mention would be useful for big companies that need a big iron box to run oracle or some other dedicated app but that's about it.

Not entirely true - an SME will look at TCO over a longer period than an enterprise company will - in which case the "consolidate and reduce TCO" cost makes even more sense. Combine that with the longevity of UNIX hardware in general, and it's a compelling play for most SMEs.

'useable container model' do you mean a java container ? the iPlanet stuff ?

No, no - OS level stuff. It's pretty interesting.

AIX has had containers for a while - however, they function more like SunFire domains - they need their own OS install etc.

With Solaris 10, Sun are dropping in a far more useful container concept - you group some resources together - CPU and memory - and then 'bind' a process to that 'container'. You can then guarantee that that process will run with those resources, independant of what happens to the rest of the domain. So you can continue to do Oracle stuff if a Weblogic app is thrashing the rest of the CPUs.

It's very very clever, and it gives Sun a very good extra point when building blade or large systems - dedicated but flexible resources like that also play in the N1 space (rolls eyes).

The interconnects would be interesting to see but would that make them a bigger player in the supercomputing space ?

Sun aren't an HPC player, and are in there more by accident than design. The interconnect thing gives more power to their product line.

SunFires are SMP machines - they have a bit of NUMA in there, but they are essentially SMP. This inhibits their scaleability.

72 CPUs in an F15k if you strip the IO boards - 512 CPUs in the Origin 3000.

A large data warehouse (say, 4 or 5 TB) is going to need as many CPUs as possible. The SMP design of the SunFires limits their expandability, which is a big selling point from Sun.

So, the Wildfire interconnect allows you to tightly couple up to 4 F15ks. It's not a single system image like an Origin - it's more an incredibly low-latency cluster like TruCluster.

The bonus is, you can use things like Oracle RAC and have much faster remote data operations across the Wildfire cluster than you would with a normal cluster. Combine that with the inherenet multi-domain design of the SunFires, and you suddenly have a very fast, very flexible and very robust solution that will scale.

That gives you a good long term platform for consolidation and reduced TCO - hence it's a winner for business.

Granted, it only really applies to the enterprise space, but E10ks and F15ks have been selling like hot cakes - there's a lot of demand for this. If you slap down £2million for a single box, you want it to last for 5-6 years before you replace it - WildFire is another way of being able to do that.


--

"Tell them we are not Gods, but Sysadmins, which is the next best thing."


[ Parent ]
The 15K is 106 CPUs. (NT) (none / 0) (#158)
by FieryTaco on Thu Aug 28, 2003 at 02:33:13 PM EST



[ Parent ]
help a newbie out? (5.00 / 1) (#128)
by el_guapo on Wed Aug 20, 2003 at 07:00:41 PM EST

what is this "kit" i keep seeing spoken of? (elguapo googles)
mas cerveza, por favor mirrors, manifestos, etc.
[ Parent ]
My guess would be... (none / 0) (#134)
by Vesperto on Wed Aug 20, 2003 at 11:48:30 PM EST

...hardware + software + assistance. Google me in on your results :)

If you disagree post, don't moderate. Alimaniere forf
[
Parent ]
um, none :-) (5.00 / 1) (#145)
by el_guapo on Thu Aug 21, 2003 at 06:06:04 PM EST

googling proved pointless. i'm a moron for not seeing that "kit" would generate, oh, exactly seven and one half infinities worth of hits :-P
mas cerveza, por favor mirrors, manifestos, etc.
[ Parent ]
kit = english slang for equipment (n/t) (none / 0) (#140)
by irixguy on Thu Aug 21, 2003 at 04:41:36 AM EST


--

"Tell them we are not Gods, but Sysadmins, which is the next best thing."


[ Parent ]
What's HP's strategy, then? (nt) (none / 0) (#116)
by nowan on Wed Aug 20, 2003 at 12:40:52 PM EST



[ Parent ]
HP's strategy (4.00 / 1) (#117)
by irixguy on Wed Aug 20, 2003 at 12:51:28 PM EST

Is, frankly, arse, and you'd be a fool to bet your business on it, but that's what vendor lock in will do - and even IBM can take lessons from HP when it comes to that.

Basically, HP have these platforms:

  • Alpha with Tru64 or VMS
  • PA-RISC with HP-UX

HP is basically gonna end up with:

  • Itanium 2 with HP-UX or VMS

Tru64 will bite the big one. TruCluster is being merged into HP-UX, which will be HP's UNIX. HP are porting HP-UX to Itanium 2, along with VMS, which they still have to support until at least 2016 (can't remember the road map). It's the legacy of all those lush government contracts DEC picked up.

Tru64 users are stuffed. They have to migrate to a new OS, and buy a new platform. Sun and SGI are offering some very sweet deals to get users to jump ship, with some good success.

HP-UX users are used to having to buy completely new hardware each upgrade cycle, so moving to Itanium 2 will come as no real shock. Plus they get decent clustering at last, in TruCluster, which is unrivalled.

HP currently have a version of HP-UX that has *some* features of TruCluster merged in, as well as some god-awful bodged up CPU modules that will let you fit Itanium 2s into a SuperDome. Seriously, it's an appalling bodge - really shocking stuff.

They also have a version of HP-UX which will run on Itanium 2, so they're getting there - but it's pretty risky, and that's frightening businesses away.


--

"Tell them we are not Gods, but Sysadmins, which is the next best thing."


[ Parent ]
Most of your facts are incorrect (5.00 / 4) (#130)
by slashcart on Wed Aug 20, 2003 at 07:45:43 PM EST

SGI never said they were transitioning to Intel CPUs ... Ever

This is completely false; SGI announced that they were ceasing development on future MIPS chips and switching to itanium. Here is a link from news.com from that era, entitled "Silicon Graphics scraps MIPS plans", that says:

"(SGI) has quietly scrapped ambitious plans for its MIPS processors... The last major MIPS release for servers and workstations is MIPS scheduled for 1999, sources said."

The E10k ... is not technologically advanced - have a look inside one. They have off-the-shelf 3COM switches inside for their interconnects.

This is totally false; the E10k does not have 3com switches for an interconnect. The E10k has a proprietary low-latency 16x16 crossbar nonblocking interconnect. Here is a link to Sun's spec page that details this.

Next, your comments about the creation of the E10k are bogus. The E10k is nothing more than a C6400 with UltraSPARC II CPUs in it.

I mentioned this in the article.

Post dot-com boom, *all* UNIX vendor's sales plummeted - this is the nature of an economic crash, and not something related soleley to Sun.

Sun's sales, stock price, and market share have plummeted enormously more than its competitors since the crash. Sun's stock price lost over 96% of its value, and even this year, Sun's hardware sales are down 20% year-over-year.

At the low end, Sun are doing fine. They have machines based on the US IIi and III which will outperform offerings from Dell and IBM

No USIIi machine would outrun any current x86 box from Dell or IBM. The fastest x86 boxes outrun the fastest US IIi boxes by more than a factor of five on SpecCPU (here, type in 'sun' in the search area). Even Pentium 3/1GHz, chips that have been long since discontinued, substantially outrun the fastest US IIi.

Sun is not "one full process generation behind in manufacturing process" - that is twaddle. What you mean to say is that their die size is not as small as Intel

This is totally false! The 1GHz US-III are manufactured using a 180nm process, while Itanium2 and POWER are manufactured using a 130nm process. (here and here) This refers to MANUFACTURING PROCESS not die size. Intel's chips have been fabbed at 130nm for almost two years! Sun's just switching to 130nm now, and Intel's switching to 90nm at the same time.

What you mean to say is that their die size is not as small as Intel

The US-III die size is much smaller than Intel's; this is largely because the US-III L2 cache is external.

The US III is not a poor performing chip, by any stretch of the imagination. It has a low clock speed and handles small integer jobs slower than an Intel CPU.

I described the performance of the US-III as "tepid" and "modest." Having a low clock speed and lower integer (and fp) scores is "tepid" and "modest."

Overall this piece is poorly researched and merely parrots the same inaccurate..

Quite the contrary, all of facts in the article were correct, and here I provided references. On the other hand, most of the facts in your response were completely incorrect.

[ Parent ]

You've still to prove that (2.00 / 1) (#139)
by irixguy on Thu Aug 21, 2003 at 04:40:33 AM EST

I'll address your points one by one:

This is totally false; the E10k does not have 3com switches for an interconnect. The E10k has a proprietary low-latency 16x16 crossbar nonblocking interconnect.

This refers to the UPA crossbar that all UltraSparcs use. This is local on the CPU boards. Look inside an E10k, and they have 3COM switches which provide the interconnect between the boards. DEC did the same thing with their Memory Channel interconnect for TruCluster - it uses a 100mb/s ethernet switch.

This is completely false; SGI announced that they were ceasing development on future MIPS chips and switching to itanium.

Untrue - you quote an article that is a journalists view of events at the time. You do not quote any SGI source saying they were planning to stop development of IRIX or their MIPS CPUs, and you never will - because they never said it. SGI's roadmap has always featured continued development of their OS and chips - it's analysts and journalists who don't understand an R&D company who assumed that SGI's ill-considered foray into the Wintel world meant they would end IRIX and MIPS development.

Sun's stock price lost over 96% of its value, and even this year, Sun's hardware sales are down 20% year-over-year.

Again, nothing you quote here contradicts anything I said, nor does it prove Sun is in trouble.

Their stock price fell because it had been over-inflated by analysts. That does not mean Sun is in trouble - it means that analysts were fools and over-hyped Sun's stock - as I said.

Sun's hardware sales fall is less than any other vendors in the enterprise UNIX space. Sure, overall IBM and HP's sales fell less - but in their UNIX divisions they took a far bigger hit than Sun. And that goes to prove my points - Sun is not in trouble, and businesses will continue to invest in Sun over other UNIX vendors because Sun have the best solutions, in that space.

The US-III die size is much smaller than Intel's; this is largely because the US-III L2 cache is external.

Totally true - I meant the manufacturing process but typed die because I was thinking about CPUs.

No USIIi machine would outrun any current x86 box from Dell or IBM.

Nonsense - you're just looking at clock speed and SpecCPU results, which are bogus, and totally meaningless. Look at throughput and IOs per second for a V480 (which is US III btw) and equivalent 4 way Wintel machines.

The V480 is a significantly faster box overall - which is what matters to a business, and is why they will buy Sun in that case.

Clock speed is utterly irrelevant in the business arena that Sun is selling in. The fact you don't seem to understand that tells me you have little real world experience in this area.

I described the performance of the US-III as "tepid" and "modest." Having a low clock speed and lower integer (and fp) scores is "tepid" and "modest."

No, it's not, and you still don't understand. It's the overall performance of the machine that matters. High clock speed is meaningless when the CPU is sitting there idle waiting for data from RAM or peripherals.

Clock speed only matters when you have small, easily paralleled jobs - like rendering. Otherwise it is a useless yard stick to measure machine performance.

Quite the contrary, all of facts in the article were correct, and here I provided references.

You've yet to refute my points - you've merely repeated your previous inaccuracies and mis-conceptions as if they were somehow gospel.


--

"Tell them we are not Gods, but Sysadmins, which is the next best thing."


[ Parent ]
Completely wrong (4.66 / 3) (#142)
by slashcart on Thu Aug 21, 2003 at 07:04:32 AM EST

The E10k has a proprietary low-latency 16x16 crossbar nonblocking interconnect.

This refers to the UPA crossbar that all UltraSparcs use. This is local on the CPU boards. Look inside an E10k, and they have 3COM switches which provide the interconnect between the boards

Again, this is completely false! Just because you saw 3com switches in an e10k cabinet does not mean they were the interconnect! How on earth would a 16x16 interconnect be local to a single board? Does that board have 16 processors and 16 banks of RAM? The E10k does not use 3com switches for the interconnect between boards. I'll quote from Sun's tech specs:

Gigaplane-XB interconnect is at the heart of the Sun Enterprise 10000 server's architecture, connecting up to 16 system boards (64 processors) ... the Gigaplane-XB interconnect provides a 16x16 non-blocking, true crossbar, connecting the system boards ... The Gigaplane-XB ... provides up to 12.8 GBps data bandwidth at a low constant latency of less than 500 ns ... Gigaplane-XB interconnect uses 12 proprietary multiplexer ASICs

In short, the e10k uses a proprietary low-latency interconnect with proprietary ASICs. Here is another link that goes into some detail about Sun's interconnect; it isn't an ethernet switch! The backplane for an E10k is a GIGAPLANE which is a proprietary low-latency interconnect running at 12.8GBps.

Untrue - you quote an article that is a journalists view of events at the time.

The journalist was quoting sources at SGI and that was made clear in the article. SGI modified its own processor roadmap presented at isscc and the h1 and h2 had been removed.

The R12k, R14k, and R16k MIPS cores were all tweaks of the R10k core introduced in 1995; these cores were designed by MIPS not SGI. The h1 and h2 were originally scheduled for release by SGI years ago; they quietly disappeared from SGI's roadmap. This is largely why there hasn't been an all-new MIPS core for years.

Sun's hardware sales fall is less than any other vendors in the enterprise UNIX space. Sure, overall IBM and HP's sales fell less - but in their UNIX divisions they took a far bigger hit than Sun.

Completely false. Sun has been rapidly losing Unix market share to both IBM and HP. In Q3 2000 (during Sun's heyday) Sun briefly had more Unix revenue than both IBM and HP combined. As of recently, however, Sun is not even if the biggest Unix supplier any more; IBM is now the biggest Unix seller (source). Granted, Sun disputes this and still claims the #1 spot, but any way you look at it, Sun and IBM are now neck-and-neck for Unix sales. This despite the fact that Sun used to boast (during its boom) that it sold five times as much unix equipment as IBM.

Look at throughput and IOs per second for a V480 ... The V480 is a significantly faster box overall ... Clock speed is utterly irrelevant in the business arena that Sun is selling in. It's the overall performance of the machine that matters. High clock speed is meaningless when the CPU is sitting there idle waiting for data from RAM or peripherals

Large-scale SMP Unix boxes are usually used for databases. Databases (OLTP) are mainly constrained by latency not bandwidth. In this regard Sun is worse than its competitors because its L2 cache is external and therefore has a much greater L2 cache latency than other systems.

Your comment about the "CPU sitting there idle" is ironic because Sun's CPU is the last in-order CPU in the entire RISC landscape, thus when Sun's chip hits a memory latency it just SITS THERE doing nothing. All the other vendors (even MIPS) managed to implement OOE in the mid-90s. Bear in mind that OOE by itself leads to ~30% greater improvement on commercial (not compute intensive) workloads.

Again, with database the primary bottleneck is latency not bandwidth. Since latency is so important to transaction-processing performance the major vendors have gone to great lengths to reduce it using complicated memory fabric designs (n-dimensional hypercube, etc) that employ proprietary ASICs. HP's orig Superdome memory fabric employed ASICs that had 40 million transistors, more than the CPUs (PA/RISC)!

This is why no vendor would ever use ethernet switches as an interconnect. Latency would be 10 milliseconds (not nanoseconds) and transaction-processing performance would be completely destroyed. And this is why there are so few vendors of large-scale SMP (>32 way) systems. If it were as easy as connecting motherboards using an ethernet switch, anyone could do it.

It's the overall performance of the machine that matters.

Sun's difficulties with it's in-order CPU and high latencies have led it to post disappointing results in application (not synthetic) benchmarks. Sun dropped out of the tpc-c benchmark when its boxes started to fall seriously behind.

Even if bandwidth is what's truly important, as you've claimed, IBM's p690 still substantially outruns an e15k in this regard.

You've yet to refute my points - you've merely repeated your previous inaccuracies
I refuted every one of your points and provided supporting documentation. Your points were:
  1. Sun e10k's use 3com ethernet switches as an inter-board interconnect
  2. The US-III is manufacted on the same process as POWER and Itanium2 (130nm).
  3. Sun has been gaining Unix marketshare from IBM and HP.

All three of these points are unquestionably false. You apparently won't believe the links I provide. Check it out by typing "sun e10000 interconnect" into google yourself.

[ Parent ]

Again, you still have yet to prove that (3.75 / 4) (#144)
by irixguy on Thu Aug 21, 2003 at 08:56:41 AM EST

Again, this is completely false! Just because you saw 3com switches in an e10k cabinet does not mean they were the interconnect! How on earth would a 16x16 interconnect be local to a single board? Does that board have 16 processors and 16 banks of RAM? The E10k does not use 3com switches for the interconnect between boards.

Eh? Do you understand what Sun actually mean when they talk about a 16x16 interconnect? It's an ASIC! It's a single chip that sits on either the CPU board, or on the backplane that it plugs into.

Does that board have 16 processors and 16 banks of RAM?

I'm sorry, but that shows a real ignorance of how this works. The capability of the cross connect has nothing to do with what is plugged in to it, other than providing a theoretical limit. Just because it's capable of 16x16 doesn't mean that's what is connected.

And what is the transport for Gigaplane? In the E10k, it is a pair of 3COM switches. Why do you think the peak theoretical bandwidth is 1.6GB/s? Full duplex GigE per chance?

I'm sorry, but it's pretty obvious you've never come near one of these machines.

Let's take the SGI Octane as an example. It's a 2 way or single CPU workstation from SGI. It's design is based on the Origin 2000. It uses Origin 2000 ASICs internally. Can you guess what the interconnect ASIC is?

Hint - it's the same 8x8 crossbar ASIC as the Origin 2000 uses.

I think this underlines the point that the capabilities of a crossbar ASIC only serve as an upper limit to what is plugged into it.

The journalist was quoting sources at SGI and that was made clear in the article.

Sources which the journalist was unable to name or quote properly. Why? Because such statements were never, ever made by SGI.

The R12k, R14k, and R16k MIPS cores were all tweaks of the R10k core introduced in 1995; these cores were designed by MIPS not SGI.

This is totally untrue. SGI designed the R10k upwards - MIPS was spun off as a seperate business a long time ago. R10k and up have been designed, in house, by SGI's engineers. SGI have not relied on MIPS since the R10k.

This is largely why there hasn't been an all-new MIPS core for years.

Have you seen SGI's roadmap? The R18k is a dual-core CPU which is undergoing trials now.

Sun is not even if the biggest Unix supplier any more; IBM is now the biggest Unix seller

Oh, come on. That IDC survey was disputed by both Sun, and HP. Quite apart from the fact that the link you provide supports what I said - all UNIX vendors took a hit, and that is not a sign that Sun itself is in decline.

Databases (OLTP) are mainly constrained by latency not bandwidth.

No, that's incorrect. If your infrastructure is designed correctly, latency doesn't even enter into it. If you database is waiting because of a slow sub-system (be it CPU, RAM, or IO) you haven't designed the infrastructure and/or the database correctly. It's as simple as that.

This is why no vendor would ever use ethernet switches as an interconnect.

You're just underlying your lack of understanding and familiarity with the technology. You have ignored my previous comment stating that Memory Channel in TruCluser uses an ethernet switch. If this is so false, and so wrong, why are HP still merging that technology into HP-UX?

Read Sun's white papers - Wildfire is based on - wait for it - GigE! Face it - vendors use ethernet switches as interconnects between nodes. Whether you think it is a good idea or not is incidental to the realities of the situation.

Latency would be 10 milliseconds (not nanoseconds) and transaction-processing performance would be completely destroyed.

Again, you show your ignorance of the technology and your lack of understanding on how it is used. Oracle RAC depends heavily on interconnect performance between nodes in a cluster. How are those cluster nodes connected? By ethernet. GigE or 100mb/s.

Yet RAC is designed for incredible performance for OLTP. Are you really so arrogant as to claim that Oracle's database development, and RAC, are flawed technologies? That RAC is just a big "smoke and mirrors" scam?

Unlike you, I have implemented RAC in an ethernet connected cluster. Performance is incredible. Latency is only an issue, as I said earlier, if your database or infrastructure design is poor.

Sun's difficulties with it's in-order CPU and high latencies have led it to post disappointing results in application (not synthetic) benchmarks. Sun dropped out of the tpc-c benchmark when its boxes started to fall seriously behind.

The tpc-c benchmark *is* a synthetic benchmark. That's why both Sun and SGI don't take part in it - it provides false, higher than real-world figures, that people then latch on to. The Origin 3000 provides the fastest, lowest latency than anything else on this planet - why on earth would SGI not want to post tpc-c benchmarks for it?

Rhetorical question. It's because it has no real world relevance. It's fine for IBM and HP to thrash around doing their little "Itanium2 vs. POWER" spat, but they are not real world results. At all. You would just never achieve that sort of performance in any installation.

Even if bandwidth is what's truly important, as you've claimed, IBM's p690 still substantially outruns an e15k in this regard.

This is completely irrelevant. I'm talking about the 4 way space, where the V480 is a leader on price/performance. If you want to quote irrelevant things, the Origin 3000 in 512 CPU config spanks the pair of them. Big deal.

1. Sun e10k's use 3com ethernet switches as an inter-board interconnect 2. The US-III is manufacted on the same process as POWER and Itanium2 (130nm). 3. Sun has been gaining Unix marketshare from IBM and HP.

Point 1 you have yet to disprove. You're merely quoted stats and examples that go into no detail about what provides the low level connection internally in an E10k. You also make it obvious you have never seen inside one, nor adminned one, making your opinions on it's construction irrelevant.

Point 2 is bogus - I admitted I had made a mistake when talking about dies. However, your original claim was that Sun were behind other vendors in CPU development - this is demonstrably false. Intel do not have a dual-core CPU on the cards for a long time. IBM have POWER4, but it's only dual. Sun not only have dual, but also quad core CPUs coming next year. They have 8 core CPUs coming the year after.

Point 3 I never claimed. I claimed that all vendors had lost revenue and sales because of the economic downturn, and that this was not exclusive to Sun, and did not show a decline on Sun's part. You have yet to demonstrate otherwise, apart from repeating analysts' flawed summaries, and quoting a disputed IDC survey.

You apparently won't believe the links I provide.

That's because the links you provide do not prove your points - at all.

So far, I'm afraid, you've demonstrated no real knowledge of any of the technologies you speak about, let alone any real world experience.

Anyone can quote analysts and throw an article together based on that, as you have done.

However, my original point still stands - this is inaccurate and poorly researched.


--

"Tell them we are not Gods, but Sysadmins, which is the next best thing."


[ Parent ]
Denial is not just a river in Egypt... (5.00 / 3) (#148)
by slashcart on Thu Aug 21, 2003 at 09:30:45 PM EST

And what is the transport for Gigaplane? In the E10k, it is a pair of 3COM switches. Why do you think the peak theoretical bandwidth is 1.6GB/s? Full duplex GigE per chance?

Here and here are two technical whitepapers written by the engineers who designed the Gigaplane backplane at Sun. These papers describe it down to the electrical level. In summary, each system board has 1,296 signal and ground pins connecting it to the backplane, for a total of 21,000 pins in a point-to-point backplane. Each board has four separate cache snoop buses, a 48-bit address bus, and a 288-bit data bus with a 12 n.s. delay.

Here are some relevant quotes from the papers:

"Each system board has 1,296 signal and ground pins, for a total of nearly 21,000 pins that go to the global interconnect.

to quadruple the snooping bandwidth, we chose to interleave four snoop buses

we used point-to-point router ASICs for all the interconnect: data, arbitration, and the four address buses. We avoided bussed signals...

An arbitration bus consisting of 16 dedicated lines, one per board, is used to implement ... algorithm.

To accomodate 16 board slots and run the bus at 12ns cycle time, the bus's phyiscal length has to be minimized. A 20.5" wide centerplane was designed such that 8 boards can be plugged from either side of the centerplane.... This low-impedance centerplane reduces bus loading delay by 46%.."

In short, Gigaplane is a proprietary low-latency point-to-point backplane with 1,296 pins for each of the 16 boards, and with unique, custom electrical design including a proprietary snooping protocol and proprietary electrical interconnects. This is detailed in the papers I provided. THIS IS NOT ETHERNET.

I'm sorry, but it's pretty obvious you've never come near one of these machines.

Dude, I've been inside almost all the high-end SMP machines. The bus interconnect on a single E10k board is about a foot wide and has over 1,000 pins, on both sides of the board. Not Ethernet. In the IBM p690 and v990 the cables for the interconnects are thicker than my arm and have thousands of lines. Not Ethernet.

You obviously saw some 3com ethernet cards in an e10k cabinet and thought that was the interconnect. I talked to a close friend of mine who was an engineer at Sun (I live in Silicon Valley) and told him about ethernet serving as the board interconnect; he laughed very hard and just about pissed in his pants.

Oracle RAC depends heavily on interconnect performance between nodes in a cluster. How are those cluster nodes connected? By ethernet. GigE or 100mb/s. Unlike you, I have implemented RAC in an ethernet connected cluster.

Oracle RAC is a cluster technology that has nothing to do with in-the-box interconnect. RAC was a technology that took billions of dollars to develop; RAC was pretty much an attempt to compensate for the incredible latencies of ethernet. This is very difficult to do and was only released somewhat recently. It has nothing to do with an e10k interconnect! A cluster is not an SMP box!

Unlike you, I have implemented RAC in an ethernet connected cluster.

Until quite recently I was serving as the Chief Database Architect for a significant corporation; I may have worked on some of the products that you deploy.

your original claim was that Sun were behind other vendors in CPU development - this is demonstrably false. Intel do not have a dual-core CPU on the cards for a long time. IBM have POWER4, but it's only dual. Sun not only have dual, but also quad core CPUs coming next year.

Demonstrably false?! The US-III is a 4-issue, 5-unit, single core, IN-ORDER, 1GHZ, 14-stage processor with external caches. This is far behind the competitors on every significant factor! The POWER4 is a 6-issue (per core), 8-unit (per core), dual core, fully out of order, 13-stage processor with twice the L1 cache, and the L2 caches integrated into the processor die! Thus, the Power4 has more units (by far), a wider issue width (by far), is fully out of order, runs at a far higher clock speed, has a much shorter pipeline, and has far bigger caches, on each of its two cores! The Itanium 2 is a 11-unit, 6-issue design; more than twice the units, and a 50% greater issue width, than the US-III. The i2, like the POWER4, manages to have a much shorter pipeline length combined with a far higher clock speed. The i2 has an 8-stage pipeline and runs at 1.5GHz, which is 40% shorter and 50% faster, respectively, than US-III. The i2 has three levels of caches on the die (compared to 1 for US-III).

Pipeline length and clock speed are inversely related design parameters and it's damn near impossible to increase clock speed at a given fab process without lengthening the pipeline. It's remarkable that both Power and i2 have far shorter pipelines and far higher clock speeds combined even when they were introduced at 180nm!

These are among the reasons that the US-III is so emarrassingly spanked on both commercial and fp workloads by i2 and POWER. i2 has over twice the floating point performance (SpecFP) as US-III! The deficiencies of Sun's UltraSparc line are widely acknowledged among chip design engineers. Here is a link to an article by a computer engineer with a page entitled "Sun: too little, but is it also too late?" Here is a quote from the article:

"it is obvious that the substantial MPU development effort Sun engages in is not delivering the goods as it should. The company stumbled badly with the UltraSPARC-III (US-III),"

You have ignored my previous comment stating that Memory Channel in TruCluser uses an ethernet switch.

TruCluster is a cluster technology which is not the same thing as a memory fabric! I politely ignored it!

Sources which the journalist was unable to name or quote properly. Why? Because such statements were never, ever made by SGI.

I'm getting tired of providing links. Just type it in to google and you'll get PAGES AND PAGES of references from highly-regarded sources (like EETimes) that quote SGI people in saying that SGI was cancelling MIPS development.

No, that's incorrect. If your infrastructure is designed correctly, latency doesn't even enter into it. If you database is waiting because of a slow sub-system (be it CPU, RAM, or IO) you haven't designed the infrastructure and/or the database correctly. It's as simple as that.

Dude, it's vastly more complicated than you know. There's a vast theoretical literature on overcoming the effects of latencies. You're an admin and plug this shit in for a living; to you it's "as simple as that."

I'm sorry, but that shows a real ignorance of how this works... You're just underlying your lack of understanding and familiarity with the technology... Again, you show your ignorance of the technology and your lack of understanding on how it is used... this is inaccurate and poorly researched...

You thought the E10k uses ethernet as a backplane and memory fabric. That's so shockingly ridiculous to anyone who actually knows how memory fabrics and backplanes are designed. Your points on the US-III were similarly absurd. I'm being as nice as I can be here. Until now I've been quite civil, but you pepper your mistaken posts with little arrogant insults!

[ Parent ]

Where Sun is definitly losing the battle (3.66 / 3) (#113)
by asad on Wed Aug 20, 2003 at 12:22:05 PM EST

In chip design, my friend runs a chip startup and not only are the linux boxes faster and cheaper but he's started to see some EDA tools become linux only.  Sun used to be what everyone was using to run their Cadence applications on but it seems that people have started to migrate CPU intensive jobs like chip design over to linux for the better price/performance ratio.
His lone Sun box has a load of 0.02, none of the developers are bothering to run anything on it.


Two different markets (4.00 / 2) (#129)
by damien on Wed Aug 20, 2003 at 07:26:51 PM EST

There are two markets for Unix servers. Sun is still well-positioned in one of these, but is in serious trouble in the other. On one hand, you have the "big iron". Some applications (databases, in particular) require a single system with many CPUs and high fault-tolerance. Fast support to fix problems is critical, as is the ability to hot-swap failed components. Sun does an excellent job in this space, and isn't in any immediate danger. On the other hand, there's the lowly 1U server. Redundancy and scalability comes from buying more machines and distributing the load across them. Vendor support becomes less important--if a machine fails, you just allocate an unused one to replace it until you can get around to fixing it. This is the space where Sun has the most to fear. Cheap Linux boxes deliver an order of magnitude more performance than Suns in the same price range. While Sun's excellent serial console support once provided a compelling argument to avoid Intel hardware, recent rack-mount PCs provide sufficient network console support to narrow the gap significantly. Sun is already beginning to lose its grip on the low-end server market. Large companies are beginning to seriously consider the costs and benefits of replacing aging Suns with new Intel-based hardware. I suspect that within five years, the once-ubiquitous 1U Sun will be considered "legacy hardware". The large databases, however, will likely still be running on the descendants of the E10k.

serial video for PCs (4.00 / 1) (#146)
by bastard operator on Thu Aug 21, 2003 at 07:24:51 PM EST

check out http://www.realweasel.com/. It lets you hit the atx reset switch remotely. it's not nearly as nice as the rsc cards sun includes in their servers but it gets the job done. the thing that is nice about sun stuff is that the hardware support is always complete. for example, there's lots of support for raid cards and whatnot in linux, openbsd and freebsd, but often the OS can't even tell when a disk fails. also, the PC world is obsessed with speed. I don't think cpu performance is the most important thing in a lot of cases in the server market. Take firewalls, name servers, mail servers, or file servers for example. I'm more interested in knowing that my OS is going to run hassle free. PC hardware has such a short lifespan. You find something that works well and then in 6 months when you need another one it's been discontinued.

[ Parent ]
nice try (4.00 / 2) (#149)
by Lin Dze on Thu Aug 21, 2003 at 09:39:34 PM EST

I think your analysis is just too simple. Sun didnt pick up the market by default and you seem to be mixing up your 'attacking' influences from MS and Intel.

Itanium was a complete flop, as you said. Itanium 2 is starting to get traction but still isnt a major player. You glossed over IBMs success with Power. Power5 is doing well and Power5+ and Power6 look to be impressive and scary, in that order.

MS seems to be moving in more due to leakage from the desktop market + "lower TCO" than through any real qualifications in the enterprise segment.
Linux has made immense gains from where it started, but I think this is due to its commodity nature than through any super technical merit. Huge gains in low-midrange servers, and specialty clusters, but not dominating the Big Iron market.
When it comes to true mission critical setups a LOT of people still love "big iron" vendors and sun especially.

-Lin Dze
"Facts don't cease to exist because they are ignored." Aldous Huxley

ok (5.00 / 1) (#150)
by slashcart on Fri Aug 22, 2003 at 12:13:13 AM EST

I think your analysis is just too simple... You glossed over IBMs success with Power. Power5 is doing well and Power5+ and Power6 look to be impressive and scary, in that order.

Perhaps I should've spent more time talking about Power 4/5/6. Clearly IBM is having significant success in developing high-end MPUs, with Power4 being impressive and Power5 even more so. But POWER is a topic all to itself. POWER4 is dual-core, Power5 is dual-core with significant SMT features. Explaining SMT, IBM's history with SMT (starting at the RSIV/64), dual cores, IBM's design approach, etc, would require a substantial treatment. The article was already overlong for something on k5 so I tried to omit things not absolutely essential for understanding the topic.

you seem to be mixing up your 'attacking' influences from MS and Intel.

I'm not sure what you mean by this.

Itanium was a complete flop, as you said. Itanium 2 is starting to get traction but still isnt a major player.

Clearly, thus far Itanium 2's influence has been largely psychological. The doubts lingering from the original Itanium have faded and everyone knows the Itanium line is for real. Itanium 2 was not strictly relevant but I wanted to include it. Since I mentioned Itanium as sucking, I didn't want to just leave it at that and give the reader an entirely false impression about the Itanium line being a disaster.

Linux has made immense gains from where it started, but I think this is due to its commodity nature than through any super technical merit. Huge gains in low-midrange servers, and specialty clusters, but not dominating the Big Iron market.

True. I tried to convey this in the article:

"Linux was regarded as highly reliable and absolutely adequate for low-end server tasks."

I could've gone into the Linux phenomenon at greater length, but of course that's a topic that could fill an entire essay (or even a book!) by itself.

I think your analysis is just too simple.

I'll admit that there are topics which I could have explored much further, for example POWER 4/5, the development of Linux, Fujitsu's Sparc boxes, the advantages of Solaris, the technical details of the e10k, etc. But unfortunately, in an article of this length one must treat things somewhat lightly. Perhaps I should've narrowed the focus a bit more for the article; the topic was very broad.

[ Parent ]

Excellent reply. (3.00 / 1) (#151)
by Lin Dze on Fri Aug 22, 2003 at 01:49:03 AM EST

This is a major area and Suns (mis)fortunes are so entwined with those of IBM, Oracle, and Intel that it would be neigh impossible to have a brief article cover everything. I simply came away fro mthe article with the impressio nthat it said "Sun was given the UNIX market, but now MS and Intel are going give them a huge fight"

I certainly think you can write but how about a series on Modern HPC History. Perhaps covering an era or company in each part. That would allow more detail and a larger overall sense of where it went and possible future directions.

PS: "The more analysts agree on some prediction, the more likely they are to be wrong."
Oh so true. Somtimes I really wonder what 'research' these analysts do. Just reading Inq or The Reg gives you much better information on mid term developments.

-Lin Dze
"Facts don't cease to exist because they are ignored." Aldous Huxley

Dammit, this is a reply to slashcart below NT (none / 0) (#152)
by Lin Dze on Fri Aug 22, 2003 at 02:02:53 AM EST

NT

-Lin Dze
"Facts don't cease to exist because they are ignored." Aldous Huxley
[ Parent ]
Obligatory SCO Comment (none / 0) (#159)
by muirhead on Mon Sep 08, 2003 at 08:02:55 AM EST

The fact that Sun are supporting SCO is a sign of the desperate state of their business.



Sun's rise and fall | 159 comments (134 topical, 25 editorial, 0 hidden)
Display: Sort:

kuro5hin.org

[XML]
All trademarks and copyrights on this page are owned by their respective companies. The Rest © 2000 - Present Kuro5hin.org Inc.
See our legalese page for copyright policies. Please also read our Privacy Policy.
Kuro5hin.org is powered by Free Software, including Apache, Perl, and Linux, The Scoop Engine that runs this site is freely available, under the terms of the GPL.
Need some help? Email help@kuro5hin.org.
My heart's the long stairs.

Powered by Scoop create account | help/FAQ | mission | links | search | IRC | YOU choose the stories!