4-Way Sun Fire V40z Reviewed

Hack Jandy writes "Anandtech has a pretty thorough analysis of Sun's V40z 4-way Opteron server that fits in a 3U. Among some of the more noteable benchmarks include a 2 minute, 30 second Linux 2.6.4 kernel compile! Who would have thought only a few years ago that Sun would be the new champion of Linux and AMD?"

Comments

Who says they are?

[IANAAC]

Who would have thought only a few years ago that Sun would be the new champion of Linux and AMD?"

They're doing what they have to do to survive.

If they had their way, it'd be Solaris/Sparc all the way.

Re:Who says they are?

[saleenS281]

you confused "survive" with "grow and maximize profit". As if Sun is going anywhere anytime soon. They're going to die just like novell, BSD, and Microsoft are.
Sun realizes that the opteron provides nearly the performance of their sparc at a cheaper price... why not bundle it up and make MORE money since the cream of the crop for them is service. And more systems sold==more people buying service contracts. And lord knows cheaper prices==more systems sold.

Re:Who says they are?

[Anonymous Coward]

nearly?

show me, dollar for dollar, which sparc out performs an opteron processor.

if you tell me "well use this sparc costing X dollars" and it's some mind numbing number...then you've failed it.

assuming that said opteron is appropriate for a given job, peformance-wise, there IS NO SPARC that could compete when cost is a consideration.

fewer and fewer computational tasks require OR CA N AFFORD 200 processor sparc boxes.

when your client base shrinks to a few first world governments and high fallutin companies.....it's time to branch out.

sparc cpu WILL go away. yea sure, eventually they will all go away...but I MEAN SOONER, rather then later.

in 5 years or less, we'll see the writing on the wall, in 8 years, sparc will be a memory.

Re:Who says they are?

[Anonymous Coward]

Do you understand written English language sentences? Are you able to comprehend paragraphs in that language also? You have no point, the differences of price was acknowledged; the very sentence "Sun realizes that the opteron provides nearly the performance of their sparc at a cheaper price" means that the performance of Sun's Sparc processor are nearly mirrored in the much cheaper Opteron from AMD. There is no contention that the Opteron is not cheaper, the comments seemingly were made because for 20 years before the Opteron (if you carry the performance differential with other microprocessors back) the expensive Sparc was the only option for that relative level of performance.

Re:Who says they are?

[Anonymous Coward]

for 20 years before the Opteron (if you carry the performance differential with other microprocessors back) the expensive Sparc was the only option for that relative level of performance.

Baloney. In terms of performance, Sparc has always trailed the competition, except possibly for a few months back in the late 80s when the SparcStation 1 pizzabox was first released (the few sparcs before that were nothing special).

Even today, Sparc trails Opteron performance. Just look up the SpecFP and SpecInt numbers, fastest opteron is faster than the fastest sparc, and that's Fujitsu's sparc, Sun's own chips are even slower.

PS, Sparc has not been around for 20 years, just barely 15.

Re:Who says they are?

[codeguy007]

Novell is dying? I think you will find that Novell is making a comeback from the dead more than dying.

Re:Who says they are?

[blasphemi]

That was his point. Read it again.

Re:Who says they are?

[Bert64]

A Sparcstation-20, which can often be acquired for free nowadays, will outperform any opteron for the same price!

Re:Who says they are?

[Bert64]

Sparc was never really designed for raw performance, but if you consider the performance drop as you increase the load on a system, sparc holds up much better than most other architectures and this is what sparc is designed for. Sparc also scales very nicely to large numbers of processors and is well proven in this field.
Also, Opteron is much newer than sparc, a lot of businesses won't trust something that hasn't been around a few years and is well proven.

Re:Who says they are?

[rainman_bc]

a lot of businesses won't trust something that hasn't been around a few years

Hmmm... When I was selling server hardware, it seemed to me that "a few" years was a long time to sacrifice performace.

A few years ago we had the P3. Are you installing P3 or Old Xeon servers from a few years ago on your network?

As well, if that was the case, we'd have been waiting until 2003 for Active Directory implementations.

Get real.

Re:Who says they are?

[Bert64]

Lots of banks still use VAX machines running VMS.. As for active directory, why? novell's directory service works better and has been around much longer. Besides, before active directory came out and novell's directory service was the only option, microsoft were telling us all we didn't need a directory service, what's changed?

Re:Who says they are?

[HuguesT]

Whoossh...

Re:Who says they are?

[HuguesT]

Hello,

*still* is the operative word here. As in "it works why change it?".

You don't see banks investing in new shiny systems for a brand new application purchasing Vax/VMS systems, even if it were possible.

Re:Who says they are?

[Bert64]

Actually, Digital tried to phase out the VAX in the early 90's when the Alpha chips came out to replace them, but Compaq sold the last VAX system in 2000, when the replacement Alpha system had already been around for 8 years. So it took 8 years for some companies to move from VAX to Alpha

Re:Who says they are?

[karakal]

> They're going to die just like novell, BSD, and > Microsoft are. IMHO MS is nowhere dying any fast soon.

Re:Who says they are?

[Octorian]

OpenBSD/sparc64 isn't a valid measure of performance. In my attempts at using it, I've noticed that OpenBSD on an UltraSPARC-IIi is SIGNIFICANTLY slower than Solaris pretty much all around. The Netra I was trying to run it on was way too sluggish. But once I dumped it in favor of Solaris 9, the machine suddenly became a lot faster.
(not to say your CPU comparison isn't accurate, but that the differences are a lot less than you make them out to be.)

Now OpenBSD/sparc (32-bit), on the other hand, tends to work very well on its intended machines.

Re:Who says they are?

[maitas]

Well.. the "opteron provides nearly the performance of their sparc at a cheaper price" part is not like that. In fact Opteron is way Faster than SPARC... Just check http://www50.sap.com/benchmark/sd2tier.asp and look for SPARC and Opteron results... You got 810 SAPs per 1.2GHz US-IV and a ProLiant DL385 got 1475 SAPs per Opteron 252...

Re:Who says they are?

[sl4shd0rk]

> nearly the performance of their sparc

I disagree. Running ps2pdf to convert a large postscript file to a pdf took me close to 30 seconds on a Blade 2000 - this was writing to local /tmp and reading from local /tmp. On an Athalon xp2000, this same file, under linux, took around 10 seconds. This is an incomplete benchmark by any means, but it raises the wtf flag. For the price of this system (~$12,000) it should blow the pants off a $300 pc system doing something like a graphics conversion. I am not comfortable siding with the RISC vs. CISC argument anymore - especially since the PC is not clocking out instructions at 8mhz anymore, and the marketing dept is always looking for that edge with the buzzwords. Sun is still making some bitchin looking cases though.

Re:Who says they are?

[IANAAC]

Maybe you equate "survive" with "die", but they are doing exactly what I said - surviving. Their profits are NOWHERE near what they were. And will probably never return to previous profitability levels.

Or maybe you just don't know their history.

In any case, I repeat: They're doing what they have to do to survive.

Re:Who says they are?

[Issue9mm]

Depends on the market. When stability is the crucial factor, and not performance, then it's not at all unlikely for that purchaser to stay behind the times.

In our environment, we have top-of-the-line, cutting edge hardware all over the place, but when we speced a box the other day for one of our help desk tools (read: not requiring high performance, but requiring relatively high availablity), we selected a box with quad Xeons.

Proven technology takes the risk out for critical apps like banking, hospitals, airlines, etc. Yeah, it's nice to be on the bleeding edge, but for applications that don't need the performance, or that can be tuned to run well on older hardware, you often see a return in uptime.

-9mm-

Re:Who says they are?

[Frumious Wombat]

Sun's had a problem for a long time. We stopped buying them when the Itanium-Is flogged the US-III for less money, then then the Itanium-2/Opterons came along and widened the gap. The slowest model of each of those was 2-3x faster for floating-point heavy simulation code, and in the case of the IA-64, would run HP-UX, if we didn't want to deal with Linux.

We told Sun about these numbers, they countered that the US-IV was on the way, and I pointed out that according to their specs, the US-IV would be no faster than the IA64s we could order then and there. The salesmen knew it, but management had made their decision "Sparc Then, Sparc Now, and Sparc Forever! (or until we have no customers, whichever comes first)"

Re:Who says they are?

[rainman_bc]

Precicisely. the OP said that we wait a few years before investing. That's bullshit.

There's a difference between replacing existing systems, and investment in new ones.

First of all, AD vs NDS isn't necessary right now. It's moot. Point is that AD adoption rate was high from the beginning.

In fact, take Intrawest, a rather large company that owns Whistler-Blackcomb. They adopted win2k3 right from the outset. I argue there's many companies that jump on bleeding edge software, and many are indifferent towards hardware in the same manner too.

Re:Who says they are?

[Anonymous Coward]

Sparc was never really designed for raw performance, but if you consider the performance drop as you increase the load on a system, sparc holds up much better than most other architectures and this is what sparc is designed for

That's a lot of hand-waving baloney. Sparc's got a whole bunch of "features" that end up actually working against that kind of scalability, like register windows for one and a historically small TLB. Stability and performance under high loads is much more a function of the system than the cpu - i.e. the i/o infrastructure, the memory subsystem and the OS itself. All of which are mostly independent of the cpu - you can put a crappy system around a sparc chip and a kickass system around an Opteron or vice versa.

Re:Who says they are?

[Anarke_Incarnate]

And quad Xeons have been around a very long time? Don't even go there with the P2 and P3 based Xeons. They don't count. They were different architecture. You are going by marketing names, not actual development and testing.

Re:Who says they are?

[rainman_bc]

So are the Xeons PII or PIII based? Intel chages the Xeon cores all the time. It's not the same processor it was "a few years ago". You're talking out your ass...

Re:Who says they are?

[Knara]

Over the head like a led zepplin.

Re:Who says they are?

[codeguy007]

He said that Sun is going to die like Novell, Microsoft and BSD are. I didn't misread it. Did you?

Re:Who says they are?

[recursiv]

For fuck's sake, let me explain this for you:

He said that Sun is going to die like Novell, Microsoft and BSD
As in, not at all. But you frequently find people saying they will die. So, to die the same way as Novell, MS, and BSD are is to not die at all.

All clear now?

Re:Who says they are?

[Mr. No Skills]

I think the term "VAX" might have been originally used (in this thread) generically. I still use the term VAX for our boxes, even though they are Alpha chip based.

As to the pace of replacement of these machines -- DEC (and others) used to make hardware with relatively conservative engineering and they lasted a long time with little attention. We still have clients running 20 year old Data General 16 bit Eclipses. As Microsoft took over the world we all got used to disposable Intel based servers with 3 year depriciation schedules.

You also aren't as wedded to the operating system anymore. Code used to be written to a compiler for that operating system. Now we can be pretty insulated from the operating system thanks to Apache, Oracle, Java, etc. Frees you up to purchase the best deal regardless of OS. No one needs to purchase a nice shiny VMS box since they can run just fine on Linux, etc.

There's little reason to not move away from Alpha now. HP has priced this stuff out of it being a rational option, and doesn't seem to be supporting some emerging standards on VMS moving forward (iSCSI is what's making me kill our VMS boxes).

Re:Who says they are?

[SunFan]

I love it (and hate it) when comments like the parent, here, get modded insightful. The SPARCstation 20 maxes out at four 200MHz Ross CPUs. It might be as fast, in aggregate, as a ~1GHz Pentium III. The SBus (like PCI) and probably the RAM in the SS20 are also comparable to a motherboard for the Pentium III. This was all very impressive for the mid-to-late 1990s, when the SS20 was hot stuff.

Re:Who says they are?

[Issue9mm]

Ignoring the baseless accusations made before the question was answered, they are, in fact, PIII Xeons, and yeah, they've been around a while.

How you think I'm "talking out of my ass" when I state that HA applications often utilize proven hardware in big business is anybody's guess. Whether or not you THINK you found a flaw in my solitary example doesn't change that it's true moreso than not.

Ciao,
-9mm-

Re:Who says they are?

[rainman_bc]

Cheers - but I'll argue many orgs don't follow your mentality.

First, Enterprise != tech world. SME accounts for something like 80% of the tech investment, and many SME businesses aren't as shy to acquire new hardware or software.

I know banking works on the "If it ain't broke don't fix it". Once in a while I sneak a peek at my bank to see what they are running. They surf using Nutscrape. Many seem to still use OS/2.

And sometimes, HA means you have to use new technology. The advances in HA hardware today is mind blowing.

Re:Who says they are?

[saleenS281]

sarcasm my good friend, you seem to not be able to detect it. None of the above are dying at all was my point :)

Re:Who says they are?

[saleenS281]

you also won't be finding an opteron for free :) I think you missed his "funny".

I suspected

[geekee]

"Who would have thought only a few years ago that Sun would be the new champion of Linux and AMD?"

I knew that the ultraSPARC was dead a few years ago. Not surprised at the current Sun situation.

Re:I suspected

[SunFan]

I knew that the ultraSPARC was dead a few years ago.

The next-generation Niagara and Rock CPUs will be SPARC. Not quite dead...

Re:I suspected

[DesScorp]

Sparc isn't dead...Sun just realized that they can't keep up with Intel and IBM in the chip wars by themselves. They've teamed up with a Japanese company (Fujitsu?) for future Sparc development. Sparc will be for high-end customers only. They're positioning Opteron for the cheap end.

Re:I suspected

[Nutria]

They're positioning Opteron for the cheap end.

They can't be so stupid as to realize that the "cheap end" gets more powerful, faster than the "high end" does.

After all, that's why the minicomputer industry was born 45 years ago, the non-hobbiest PC 25 years ago and Sun killed the minicomputer 15 years ago.

Re:I suspected

[k8to]

Well it seems true that currently opteron chips are not well set up for medium scale parallel processing in the style of starfires etc. They'll probably get there, however.

Re:I suspected

[codeguy007]

Current Sparcs are slower than Opterons. Sun have been running on the fumes of their installed customer base for years now. If they weren't the CPU of choice in the Oil patch they would be dead now.

Re:I suspected

[Alioth]

That seems like a really ropy strategy - Sparc on the high end and Opteron at the low end means the high and low end will NOT be binary compatible.

And people whine enough about the minor binary incompatibilities between different Linux distros due to library differences. This means the high end stuff can only run programs for the low end stuff under emulation, and vice versa.

If all the important software for Sun's corporate clients was open source, this may be less of a problem - but it isn't. It'll increase the costs to vendors having to now support two architectures for Sun, and so they'll probably drop one of the architectures.

Re:I suspected

[vrai]

Can you fit more than eight Opterons in a single machine? Can the CPUs be hot swapped? Do they have the proven uptime record of UltraSparcs?

If the answer to any of these questions is 'No' then I forsee a continued market for Sparc hardware. Banks spend millions on new Sparc kit every year - for both new and legacy applications. Contrary to popular Slashdot belief, not every task is suitable for clustering. The bandwidth between nodes is still far too small, and the network induced lag far too great.

When you can get five-nines uptime out of a thirty processor Opteron box - then it'll be time to retire the Sparc range. Until that day comes they'll always have a market.

Re:I suspected

[vrai]

Do I care? A Sun machine with the uptime promise like that costs over a million dollars.

Guess what? Not all companies require the same level of reliability out of their machines! For you reliability isn't the be all and end all. Cool, don't spend big bucks on a Sun. For many banks reliability comes ahead of everything - down time can be hugely expensive.

Oh, and a million dollars is not a lot of money to spend on a good computer. Even my small department spent more than that on Sun gear in 2004.

Written a different way most applications can be parallelized.

Most, but not all - and the ones that can't are often very important to companies.

Sun (and other big iron providers like IBM) fill a niche in the market that commodity hardware cannot fill. It is however a niche that many, very wealthy, companies need filling and one that isn't going to go away soon.

Re:I suspected

[chris_mahan]

Google gets better than 5 nines.
They don't run sparc machines, but lots of cheaper ones.
What this I hear? Could it be the sound of 100,000 servers? All hot-swappable?

When you can get a complete server, with ram, nic, cpu and disk-storage, for less than a sparc cpu, my friend, you don't cluster cpus, you cluster servers.

Re:I suspected

[georgewilliamherbert]

Google gets better than 5 nines. They don't run sparc machines, but lots of cheaper ones. What this I hear? Could it be the sound of 100,000 servers? All hot-swappable? When you can get a complete server, with ram, nic, cpu and disk-storage, for less than a sparc cpu, my friend, you don't cluster cpus, you cluster servers.

100,000 servers and $100 million bucks in custom software or more.

If your application is embarrassingly multi-machine parallelizable then you'd be a fool not to use stacks-o-workstations.

Every "big" machine that Sun, HP, IBM, etc sell is a calculated analysis by its purchasers that their application isn't embarrassingly multi-machine parallelizable. Everyone is aware of large Linux or (x)BSD or whatnot clusters. Many people whose apps are parallelizable are buying them in bulk. But they don't answer every price/performance/reliability question.

In many cases, they cost more, are less reliable, take up more server room space and power than the alternative single larger SMP box. Even today.

Re:I suspected

[Nutria]

Quite thinking like it 1990. If you have a data center that is dependent upon one machine being up 24 hours a day then you have a data security problem. The problem is your planning. All your eggs are in one basket. Any real data center will have at least 3 machines so one can be taken offline and the other can continue to do business. In fact, it is counter productive to have a power machine so tied to one infrastructure/building. You want less powerful machines spread out across a large local. Therfore, your argument is less relevant about the facts of integrity and more about people buying in stupidity.

Your arguments are the same made by DEC during the 1980s, regarding VAXclusters.

They are still valid with VMSclusters.

4-ways are nice and all

[Shut the fuck up!]

But for years I have been looking for a 3-way. My wife, is uh, not very compatible.

Re:4-ways are nice and all

[woah]

But 4-way is often 2 chicks and 2 dudes, which means being in close proximity to another dudes naked butt. Are you sure you're ready for such a big step?

Re:4-ways are nice and all

[BooRadley]

I'm sure she'll be just fine with it if she gets to pick the other guy.

Re:4-ways are nice and all

[sysadmn]

Hot pix of a nice 3-way. There's all kinds of good stuff here.

Re:4-ways are nice and all

[ArsonSmith]

Don't ruin the fantasy.

Re:I love the combination...

[GreyWolf3000]

Ummm...what do you think Gentoo is? A *BSD distribution?

At any rate, this is supposed to be a server. This thing could handle lots of SQL transactions, send and receive mail, serve webpages, and even, as you might have guessed, compile stuff. All of these can be done on any distribution.

Solaris and AMD

[uid100]

I have been running six V20z in production for about ten months now. They have - and will continue to - run Solaris. These servers have been as stable and predictiable as the V480's I manage, but compile Apache in 1/5 the time. They are definitly a sweet hardware platform, but why discount Solaris on them (in the title of this "news", by omition?)

My new AMD64 powered Gateway 7405GX is running Solaris-10 - works great! And a 64 bit kernel.

Re:Solaris and AMD

[Anonymous Coward]

My Super Mario 64 has a 64 bit kernel too
big deal

Re:Solaris and AMD

[Jeff.Schramm]

I hadn't noticed the Solaris Linux Application Environment in Solaris 10 before. Sun has some interesting things to say about it at http://www.sun.com/software/solaris/ds/linux_inter op.jsp. I wonder how well it works.

Sun is taking Linux interoperability to the next level with the new Solaris Linux Application Environment feature in the Solaris 10 Operating System (OS) for AMD Opteron and Intel Xeon x86-based systems. The Solaris Linux Application Environment (LAE) allows Linux applications to run unchanged on the Solaris OS when coupled with a Linux distribution. This enables businesses to take advantage of the innovations in the Solaris 10 OS without sacrificing investments in existing Linux applications.

Re:Solaris and AMD

[RLiegh]

>They are definitly a sweet hardware platform, but why discount Solaris on them (in the title of this "news", by omition?)

I think that when you talk about running other architectures, people are more used to thinking of NetBSD or maybe Linux, both of which have been ported to more platforms than Solaris has (afaik).

Re:Solaris and AMD

[ArbitraryConstant]

I agree.

I wouldn't want Red Hat or Suse or Gentoo on a production server, but I'd be happy with FreeBSD or Debian.

But I'd also be happy to run Solaris though. It has features that Linux and the BSDs don't have. Doesn't make it better for everything, but it's certainly worth looking at.

Re:Solaris and AMD

[rs79]

"I wouldn't want Red Hat or Suse or Gentoo on a production server, but I'd be happy with FreeBSD or Debian.

But I'd also be happy to run Solaris though. It has features that Linux and the BSDs don't have. Doesn't make it better for everything, but it's certainly worth looking at."

What he said. I find Debian annoying compared to BSD or Solaris but that may be personal preference.

Re:Solaris and AMD

[k8to]

Strange.

I've actually _tried_ all those distributions on a Sun 20z, and while Gentoo and SuSE both worked fine, FreeBSD and Debian aren't even ready for x86_64. Red Hat was notably unworkable, sadly. Maybe we only sacrificed enough goats for two distritubions.

Re:Solaris and AMD

[ArbitraryConstant]

"I've actually _tried_ all those distributions on a Sun 20z, and while Gentoo and SuSE both worked fine,"

From experience, I don't really put too much stock in the Suse or Gentoo definition of "working fine".

"FreeBSD and Debian aren't even ready for x86_64. Red Hat was notably unworkable, sadly. Maybe we only sacrificed enough goats for two distritubions."

Unless you specifically need 64 bits, you can run the 32-bit x86 port until the 64-bit port matures.

All the 64-bit ports are a bit young at the moment.

Re:Solaris and AMD

[ArbitraryConstant]

"I find Debian annoying compared to BSD or Solaris but that may be personal preference."

Some of the automation in Debian makes me nervous, but I agree that it's a personal preference thing. I probably don't use it enough to be 100% comfortable.

Re:Solaris and AMD

[codeguy007]

BSD initscripts are such a pain in the ass. System V initscripts are some much more logical and easy to work with.

Re:Solaris and AMD

[luvirini]

It seems to work very well. I was actually surpised. A customer asked to run our application on solaris and wanted us to test it. We were able to run the linux version with no changes on solaris and thus the customer went with that solution.

Re:Solaris and AMD

[flynn_nrg]

Au contraire, NetBSD-style init scripts are much more logical than sys V. With sys V you have a mess of symlinks. On NetBSD (and FreeBSD 5.x) you tell the rcorder program what dependencies your script has and what does it provide, the system figures our the execution order on it's own. To turn a service on you just put in your /etc/rc.conf file. Want to run bind? Just put named_enable="YES" in your /etc/rc.conf and there you go. Check Luke Mewburn's rc.d paper for more info.

Re:Solaris and AMD

[jpc]

Whats the point in buying a 64 bit system if you run it in 32 bit mode. The Debian approach (lets argue and not get this right for too long and not really get our act together) has been a disaster. Fedora since FC1 release has been pretty good (just upgraded to FC3), just a few teething troubles. Gentoo is apparently pretty good too, intend to try it soon.

Re:Solaris and AMD

[sunwukong]

My only complaint about Gentoo (compared w/ say, FreeBSD or OpenBSD) from an admin point of view on the AMD64 platform is that you've really got to keep up with the postings from the engineering team.

I've been running for almost a year now and while the architectural changes haven't been crippling, they weren't trivial either.

Re:Solaris and AMD

[ArbitraryConstant]

"Whats the point in buying a 64 bit system if you run it in 32 bit mode."

'Cause it's still a really nice machine?

"The Debian approach (lets argue and not get this right for too long and not really get our act together) has been a disaster."

The advantage, though, is that a stable release is very stable.

"Gentoo is apparently pretty good too, intend to try it soon."

They change too much stuff without enough testing for a production system, IMO. I've always had regular breakages when I've used Gentoo.

Some people want features and new versions. Gentoo is good for this, but it has a lot of troubles from staying that up to date, because the maintainers can't do proper regression testing every time something updates. Reliability suffers a lot, and drives people like me straight into Debian.

My motivation for using Debian comes from using Gentoo. I found the unreliability so profoundly annoying that I no longer care about using the most up to date version.

Re:Solaris and AMD

[codeguy007]

Yeah that's a newer implementation. I was talking about BSD initscripts. It's nice that NetBSD/FreeBSD has updated them to a more logical form but they are no longer true BSD initscripts then are they.

Re:Solaris and AMD

[k8to]

This is pretty foolish. If you're going to run x86, there are a lot of cheaper machines that work just fine and will give similar performance. Running x86 on amd64 is basically throwing 40% of your performance out the window, as well as memory advantages, and running a needlessly power-hungry (read expensive to power) machine.

Rebadged Newisys 4300?

[Master Bait]

Not that that is a bad thing, but I cannot see any difference between the V40z and this.

Re:Rebadged Newisys 4300?

[NaruVonWilkins]

Mod parent up. I've been using one of these as a footrest at work for over a year (albeit with slower chips).

Re:Rebadged Newisys 4300?

[Mistah Blue]

That is really interesting. I thought Sun was building these themselves. Perhaps I'm mistaken.

Re:Rebadged Newisys 4300?

[ArkiMage]

Yeah.. We have multiple Newisys 2100's and recently bought Sun 20z's which are the same thing. Cheaper as Sun than Newisys as well. The 4300 and 40z are identical as well. Oh, and Newisys will NOT offer firmware on their website. Sun does...

Re:Rebadged Newisys 4300?

[jpmoney]

Support, support, support...

At least they are all much better than the APPRO 4-ways (and the other copies of the AMD reference platform).

Yes, we have a V40z at work, so I'm a bit biased...

Re:Rebadged Newisys 4300?

[Anonymous Coward]

That's because there is none.

Re:Rebadged Newisys 4300?

[voidptr]

Silicon Mechanics sells the same box as well under the nServ A400 name.

Re:Rebadged Newisys 4300?

[Frumious Wombat]

Interestingly, if you remove the white "do not remove" stickers on the motherboard of their Opteron workstation, you'll find underneath three little letters, "IBM".

The question then starts to become, "why don't I buy an IBM/IBM workstation, or an HP-4way Opteron, and run Solaris-10 on it?"

Hmmm

[Neil Blender]

Reading through the benchmarks, I see they compiled KDE under gentoo in just under 17 weeks. I'm impressed.

Re:Hmmm

[jd]

You think that's bad? You notice the benchmark is for 2.6.4, but kernel.org is at 2.6.11-pre4-mm1. That tells you how long it took to do the download on this box.

Fast Kernel Compile

[AFCArchvile]

2 minute, 30 second Linux 2.6.4 kernel compile!

That's pretty fast compared to what I've done: compiling 2.4.27 in Gentoo on a Sun Ultra 2 (2 x 300 MHz UltraSPARC). It took over 90 minutes, and that was without the USB and Bluetooth sections of the kernel, since there's no way the Ultra 2 can make any use of either.

Re:Fast Kernel Compile

[AFCArchvile]

Yeah, I had to devote an entire day to installing Gentoo, and the box didn't have an Internet connection. I had to boot the Gentoo CD with the "docache" option at the SILO prompt to allow the ejecting of the CDROM, and dump the entire package CD onto the hard drive. Even then I couldn't get a number of packages.

The box isn't running Gentoo normally, though; I basically wanted to go through a Gentoo install successfully so I could do it all again in less time when I have the box situated in a semi-permanent, usable place. It's running Solaris 10 now, which is pretty cool, but Java Desktop runs pretty slow. Still very usable in CDE though. This computer isn't used regularly; I just do my *nix experimentation on it now and then.

Some may ask "Why not use Debian instead if you have no Internet?" I already tried; Debian Woody doesn't interface with the hardware RTC chip or the power interface, so I couldn't reliably change the system clock, and I'd have to manually shut down the box. I have the same Debian running in a Sparcstation 5 though, and it doesn't have any those incompatibility problems.

Re:Fast Kernel Compile

[incabulos]

There was a /. article in the last year or two about compiling the kernel ( may have been a 2.4.x series build ) on a 72-way E15000 IIRC. It was done in less than a minute, but having a commodity piece of hardware available today that is nearly comparible in this type of performance category at an inexpensive price is an amazing amount of progress.

Though its an UltraSPARC III vs Opteron comparison, so comparing the technology requires a bit more abstraction than a strictly x86-ISA-only set of cpus. Still, you have to admit that Sun will like winning performance prizes, its a long long time since the old but reliable ultrasparc was seen as a speed demon

Sun was very quick to announce the next generation V40z (4 x Opteron 852, 8GB PC-3200) that set more than half a dozen performance records at LinuxWorld last week.

Sun also have the edge in that they handle 8 way, 16 way, 32 way, and larger systems regularly. The design experience under their belt will help them bring product to market that utilises similar number of opterons in a single system. HP and IBM are really the only vendors with a hope of bringing out similar hardware, and Sun has been doing it better than either of them for much longer. Sun also are more committed to AMD than Intel, whereas HP & IBM are hedging their bets to some extent. Especially HP, who are still largely operating in Itanium-advocacy mode.

Than again, Suns finances are woeful at the moment, which levels the field a bit as they cant afford extravagant R&D budgets.

Re:Fast Kernel Compile

[jrockway]

All Linux distros are the same. If something works in Gentoo and not Debian, it's because Gentoo set it up and Debian didn't. If you knew what you were doing, though, you could get it working in Debian, because Debian is the same OS/drivers as Gentoo (with a different name and init scripts).

Anyway, the solution to your problem is:

# modprobe rtc
# date -s time
# hwclock --systohc

Good luck. (And I hear Ubuntu is a good Debian-based distro.)

Re:Fast Kernel Compile

[Bert64]

HP did have experience, the DEC Alpha Wildfire systems scale closer to linearly than anything else i've seen, and can go upto 128 cpus.. Unfortunately, HP decided to drop the Alpha and most of the good engineers migrated away.

Re:Fast Kernel Compile

[adam613]

I also have an Ultra 2 (2x300) that I installed Gentoo on. It took two days over a serial console (18 hours of compilation).

The real problem wasn't the speed of the CPUs; the disk performance on this thing is ABYSMAL. As an experiment, I partitioned an SATA hard drive on my amd64 machine and mounted the U2's /usr, /var, and /opt partitions over nfs (after building a LiveCD that could do so). The compile time dropped to 14 hours.

Crazy stuff.

Re:Fast Kernel Compile

[LuSiDe]

I don't get it why you compare these 2 systems. Your compare is not based on either equal price or equal performance. Instead, its an old model Sun workstation system which is compared to a new model Sun server. Different cost, different performance, different age, different architecture, different purpose. About the only similarity is that they're both from Sun.

Champion of Linux?

[Anonymous Coward]

Maybe if Sun hadn't given a ton of money to SCO, but they did so no, Sun is not a champion of Linux.

Re:Champion of Linux?

[nonmaskable]

In the "driver deal", SCO gave Sun options to buy a bunch of SCO stock at $1.83 per share. This was shortly before SCO's big linux license FUD PR campaign started, and the stock went over $20/share shortly thereafter. Combine that with McNealy gloating about having the only legal version of linux right after the "driver deal" and there's a lot more here than a conspiracy theory.

Re:Champion of Linux?

[SunFan]

Combine that with McNealy gloating about having the only legal version of linux right after the "driver deal" and there's a lot more here than a conspiracy theory.

Sun does provide a level of indemnification that no other business can. This is a strategic advantage, so, at worst, Sun is not against Linux but enjoyed jabbing IBM in the ribs a few times.
I can see that (keep Linux around for everyone, just make IBM uncomfortable for a while). That would be business as usual in the computer world.

Specs

[zerocool^]

From Sun's site:

http://www.sun.com/servers/entry/v40z/index.jsp

* Linear Processor Scalability
* Lights Out Management (LOM) with integrated service processor
* Redundant, hot-swap power and cooling
* Supports existing 32-bit x86 OS and applications

* Up to 4 AMD Opteron 800 Series processors
* Up to 32 GB
* Up to six hot-swap Ultra320 SCSI disks

- Solaris 10 on x64
- Solaris 9 HW 4/04 OS or later for x86 Platforms
- Red Hat Enterprise Linux 3 for AMD Opteron
- SUSE Linux Enterprise Server 8
- SUSE Linux Enterprise Server 9
- SUSE Linux 9 Professional (Community Edition)
- Microsoft Windows 2000 (WHCL-certified)
- Microsoft Windows Server 2003 (WHCL-certified)

The price, listed at http://www.sun.com/emrkt/opteronpromo/product.html
shows the server @ $5945, which imho is quite a reasonable price for this kind of heavy hitting hardware.

I've always had a thing for sun hardware. It's just... sexy.

~Wx

Re:Specs

[Neil Blender]

shows the server @ $5945, which imho is quite a reasonable price for this kind of heavy hitting hardware.

Not that I looked or anything, but I am sure $5945 most likely gets you 1 weak processor, the onboard ram and an ide drive. Max it out and you could be looking an $20K or more.

Re:Specs

[AFCArchvile]

That specsheet says "2 Model 844", which would imply only two Opterons. While that's not too shabby, this is a quad CPU server; why not smoke 'em if you got 'em (or just go buy the V20z)? I haven't read through the whole article, so I don't know what the $20K reviewing sample has in it, but it still sounds enticing.

Re:Specs

[Anonymous Coward]

The link that grandparent provides has the following specs:

Sun Fire V40z Server Special
2 AMD Opteron Model 844 Processors
2-GB Memory
1 73-GB 10000 RPM Ultra320 SCSI Disk Drive
1 DVD-ROM/Floppy Drive
2 10/100/1000 Ethernet Ports
1 USB Port
1 Serial Port
Lights Out Management (LOM) Software
Solaris 9 4/04 x86 Operating System

Doesn't seem too bad, considering they charge an extra $3300.00 (US) per processor. After adding 2 more processors, 4GB RAM, (2) 146GB Hard Drives, and upgrading the warranty to 3 Years Silver Support, the total went all the way up to $17,620.00.

Re:Specs

[BCW2]

That price is lowball, not fully equiped. In serious drool mode I priced out a Tyan K8Q type mobo (quad Opteron), 4 Opteron 850's, and 32 gigs of memory. On pricewatch thats over $12,000. You still need a case, PS, drives,... and thats not from vendors I would buy from based on their ratings either. Reputable vendors have higher prices. Sun of course gets the volume discount if this takes off for them.

Re:Specs

[dbIII]

I've always had a thing for sun hardware. It's just... sexy.

Here's a case mod idea - bring pizzas to geekfests in sparcstation cases.

Re:Specs

[codeguy007]

That's expensive. We sell our 4U quads for less than that. 3300 US per proc is way high btw. AMD's price for new 852 is only $1514

Re:Specs

[quarkscat]

Yawn!

I think I'll just wait until the dual core
Opteron is available as an option. The only
thing better than a 4x 64-bit processor setup
in a 3U chassis is a 8x setup (unless you're
paying Oracle's per processor tax).

Re:Specs

[4of12]

shows the server @ $5945, which imho is quite a reasonable price for this kind of heavy hitting hardware.

I've always had a thing for sun hardware. It's just... sexy.

Sun hardware has traditionally been through a more rigorous design and testing phase than most PC hardware.

I have to wonder if those quality standards have been maintained on their Opteron system.

For eons, AMD has been trying to overcome an image of offering cheaper x86 hardware that is built into shoddy white boxes with bad power supplies and fans that whine at a premature age.

Sun's high quality hardware reputation would be a great way for AMD to make inroads into the lucrative corporate market. But AMD could sabotage that improved reputation for quality by selling Opterons to system builders that are willing to cut corners that would cast their product in a poor light.

Re:I love the combination...

[tealtalon]

I wonder how long to kompile KDE. That's the worst part of a Gentoo install for me.

Re:What happens when you try to install windows on

[SunFan]

Sun won't sell you Windows, but these servers are "Windows Certified". This is just fine if your company happens to have a site license.

Is 150 second Kernel Compile really that fast?

[GGardner]

OK, it's a lot faster than my machine. But how many lines of code is that? And how many lines per second? I know that gcc isn't optimized for compiling speed, far from it, but shouldn't we be getting 100klines per second out of our compilers these days?

Re:Is 150 second Kernel Compile really that fast?

[Monkelectric]

I really dont think it is. My Athlon 1700 compiles a kernel in under 5 mins. I think my P4 is 3 - 4 minutes.

I suspect that kernel building does not run in parallel very easy.

Re:Is 150 second Kernel Compile really that fast?

[timeOday]

I suspect that kernel building does not run in parallel very easy.

The kneejerk response here is, "just use make -j 10" But in practice, I don't think make will parallel compile files in different directories, and linking is a serial bottleneck. In a kernel tree I expect linking is performed in many different directories to produce modules.

Re:Is 150 second Kernel Compile really that fast?

[davecb]

Consider Amdahl's law: the compilation parallelizes well, yielding .o's, followed by a big, slow, single-threaded link., which defines the lower bound for how long the whole job takes.

--dave

Sybase ASE 64-bit Opteron?

[rngadam]

Grrr, any Sybase engineer could tell when the HELL they are going to deliver Sybase ASE on Linux 64-bit for Opteron???

We're just waiting for this at work to move to all this cool hardware! Geez... chalk one more for moving to Oracle!

Re:Sybase ASE 64-bit Opteron?

[kaiwai]

Why use Linux? I mean, not to sound like a flamer, but if you so want Sybase that badly, then go out, install Solaris 10, buy a license, which is only a few hundred a year, and throw on 64bit Sybase.

Sure, it doesn't have the "Linux is cool" factor, but you're running a computer for a business, coolness shouldn't even enter the equation. Does it work? does it run on the smell of an oily rag? is it stable? is it reliable?

Re:Since when...

[Lehk228]

Since when... (Score:1)
the kernel compilarion speed is a benchmark factor for a server hardware.

because it is something that many home users as well as server admins have actually performed on various machines and gives a better measure of performance to people than some arbitrary benchmark score.

Re:Curious

[caino59]

its free marketing.

the purpose here is too move product after all and make their shareholders money.

im not surprised at all

Re:I love the combination...

[TheKarateMaster]

Yeah... umm... I'm *pretty* sure Gentoo *is* Linux...

Don't want to start a disto war here or anything, but I think we can all agree that Linspire is by *far* the best, anyway.

P.S.(Don't shoot me, I'm just kidding, I use Debian.)

4-year-old dupe :)

[moosesocks]

4 years ago slashdot posted a story introducing the first Dual-processor athlon system and used the linux kernel compilation time as a benchmark.

A little over 4 years ago, a Dual Processor Athlon System compiled the kernel in 2 minutes flat. The kernel was version 2.4.0ac12.

I'm no software/hardware developer, so I'm not going to comment on the significance of this result, but nonetheless I find it interesting that the kernel took less time to compile on a much more modest system 4 years ago. Has the kernel really grown THAT much?

Think about it --- they were using two 1.2ghz 32-bit processors with 256mb of ram opposed to the four 64-bit processors with 8gb of ram in this test, and it still took 20% longer to compile!!!

Re:4-year-old dupe :)

[fitten]

Welcome to software... coders can't resist adding that *one more feature*. Some features are nice, others aren't so much. The Linux in-memory image (including X because Windows has a GUI) is more resource intensive than Windows these days.

Re:4-year-old dupe :)

[BillKaos]

Check the compiler version. New gcc versions got very slow.

Andrew Morton uses gcc 2.95, because it's 2x faster compiling the kernel.

Re:4-year-old dupe :)

[Eil]

Has the kernel really grown THAT much?

Yes and no... Using kernel compile times as a benchmark is categorically useless you quote the exact config file in the analysis.

A few weeks ago, I tried to compile a GNU/Debian Linux 2.6.x kernel on a Pentium III using the default kitchen-sink config. After about an hour and a half of just sitting there waiting for the damn thing to finish (this was on-site maintenance of a critical mail server), I halted the build and took my chances at configuring it by hand, hoping that I wouldn't forget some option that caused the machine not to boot. After paring it down, the new kernel plus a few modules were fully built in less than 10 minutes. (And it even booted fine.)

Re:4-year-old dupe :)

[zyridium]

That sounds really clever. Because the only thing I care about with an ahead of time optimizing compiler is the compile time.

*Maybe* that 2x is giving a couple percent of additional performance...

Re:4-year-old dupe :)

[Bert64]

Yes, just when computers got fast enough to compile 2.0.x quickly, we got 2.2.x.. But that's always the case...
Software gets bigger and slower while the hardware gets faster, so that the overall user experience remains the same.

Re:4-year-old dupe :)

[Bert64]

Well that's because ms stop releasing security patches for old versions, and old versions aren't extensible nor do they support modern hardware, finally prior to win2k, windows was unuseably unstable with the exception of NT3 so there was little incentive to use an old version.
In contrast, linux kernel 2.0.x is still maintained, and will run on some most hardware, and if it doesn't your free to port drivers to it yourself or hire someone else to do it, it's also very stable.

Re:4-year-old dupe :)

[isorox]

The Linux in-memory image (including X because Windows has a GUI)

Windows might have a GUI, but what good is a GUI when you have no monitor?

Re:I love the combination...

[TheKarateMaster]

Maybe it's a sign... don't use KDE.

Ah, it's not that bad, I'm just jealous because I've only got a 333MHz, which can *not* run KDE. (Well, it *can*, but only if you like your browser to take 45 seconds to open :P)

2.4/2.6 compile times compared -- v/s whitebox

[unixwin]

Before the Sun lovers go chanting ga-ga-ga about how this will save Sun's sorry ass or how it outperforms their "other" systems , I'd like to put forward some numbers running similar tests against whitebox systems.

Config: On my 8GB 246 (single processor, whitebox) opteron I get (make distclean etc between steps)
Time / Kernel / Make option
2"12s / 2.4.21 (time make -j5)
3m33.081s / 2.6.4 (time make -j5)
3m31s / 2.6.4 (time make -3)

From anandtech for the 2.6.4 kernel.
2"43 sec V40Z -j5
3"30 sec V40z -j3
4" 34 sec W2100Z -j3

Hmm.. for the 5K I paid for it. I'm happy waiting 50 seconds more.. ( 5K v/s 17K and 3"30' v/s 2"43')

Misc info:..
gcc -v
Reading specs from /usr/lib/gcc-lib/x86_64-redhat-linux/3.2.3/specs
Configured with: ../configure --prefix=/usr --mandir=/usr/share/man --infodir=/usr/share/info --enable-shared --enable-threads=posix --disable-checking --with-system-zlib --enable-__cxa_atexit --host=x86_64-redhat-linux
Thread model: posix
gcc version 3.2.3 20030502 (Red Hat Linux 3.2.3-42)

make --version
GNU Make version 3.79.1, by Richard Stallman and Roland McGrath.
Built for x86_64-redhat-linux-gnu

Re:2.4/2.6 compile times compared -- v/s whitebox

[byronmiller]

Compiling is hardly a server benchmark. What you need to look at is benchmarks in regards to transactions per second in an RDBMS, JVM, Apache server or Mysql and such. The real beauty of a solid engineered system is the stability, bandwidth and capacity. Try running Oracle 10g on your whitebox as spec'd with 500 concurrent users doing OLTP and let me know which one works (and doesn't stop working) the best.

Re:2.4/2.6 compile times compared -- v/s whitebox

[Cthefuture]

Your comparison is invalid without knowing the kernel options they had enabled. You would need to set the exact same configuration to make any comparisons because disabling options makes the kernel compile faster (a whole lot faster in some cases).

In fact, your times seem impossible. I would like to know your kernel options because I just compiled a 2.6.4 kernel in 6m 21s. This is on an Opteron 250 (UP) machine with my normal kernel options.

Re:2.4/2.6 compile times compared -- v/s whitebox

[unixwin]

cd to the directory with the kernel src
make clean / make distclean
yes "" | make config
make -jX

exactly as described in the article.
There is no build modules stage in this (or atleast I didn't do that)

Re:2.4/2.6 compile times compared -- v/s whitebox

[Cthefuture]

Yeah, sorry about that, I didn't see the # yes "" | make config thing. I recompiled with those settings in 6m 34s.

There is no way your 246 compiles that fast. How can that be?

Re:2.4/2.6 compile times compared -- v/s whitebox

[prockcore]

There is no way your 246 compiles that fast. How can that be?

Easy.. caching. He compiled it at least twice in a row. I bet if he did a clean boot it would take a lot longer.

Re:2.4/2.6 compile times compared -- v/s whitebox

[Cthefuture]

No, that wouldn't make that much difference.

My 2.4.22 compile time seems a lot closer at 2m 2s.

His 2.6 compile times seem impossible unless he is using 2 processors.

Re:2.4/2.6 compile times compared -- v/s whitebox

[Cthefuture]

Ah, I just looked at the Anandtech graphs. They have the V40z compile with -j1 (ie. only using one processor) at over 10 minutes. That's more like it (actually, seems a little long).

There is no freaking way a single 246 can compile the kernel in 3.5 minutes. Impossible.

Re:2.4/2.6 compile times compared -- v/s whitebox

[pavera]

you are completely full of it.
Anand took over 10 minutes with j1, since you have a single proc, doing j3, j5 whatever will make the build slower as you'll lose cycles in process switching, you won't gain anything by running multiple jobs because you have only 1 processor. It is 100% impossible for you to have built the kernel with the same options any faster than their -j1 build.

Re:2.4/2.6 compile times compared -- v/s whitebox

[unixwin]

-j1 does not mean using one processor. It means using one job. -j5 makes (at max) 5 jobs run , it does not magically increase the number of processors you have.

-j jobs
Specifies the number of jobs (commands) to run simultaneously. If there is more than one -j option, the last one is effective. If the -j option is given without an argument, make will not limit the number of jobs that can run simultaneously.

Regarding the time of 3.5 mins. I rechecked it again. I re-downloaded 2.6.4.tar.gz from kernel.org again,untar etc.
make distclean
yes ""|make ;time make -j3
again the numbers are:

Root device is (8, 1)
Boot sector 512 bytes.
Setup is 2621 bytes.
System is 1706 kB
Kernel: arch/x86_64/boot/bzImage is ready

real 3m31.077s
user 3m10.780s
sys 0m20.050s

The only thing I can think of is that perhaps you are using a different 2.6 kernel ?? Why don't you try doing the same process yourself instead of believing what anandtech is telling you??

Re:2.4/2.6 compile times compared -- v/s whitebox

[Cthefuture]

-j1 does not mean using one processor. It means using one job. -j5 makes (at max) 5 jobs run , it does not magically increase the number of processors you have.

Well duh. -j1 means only one process will run at a time. What this means is that on a machine with multiple CPU's only one CPU will be used at a time. In other words it's the same as compiling on a single processor machine. For example, if I have two identical machines, one has a single processor and the other has multiple processors. If I do a "make -j1" on both machines they will take exactly the same amount of time no matter how many processors the machines have. The V40z with -j1 took 10 minutes to compile the kernel (ie. using only one 850 processor).

Why don't you try doing the same process yourself instead of believing what anandtech is telling you??

I did do the process. See all those compilation times I posted before? That's for my single processor 250 machine. 6.5 minutes. Unfortunately I don't have a V40z to test on so I can't recreate everything.

I suspect the difference in your system is the fact you are compiling in 64-bit mode where I am compiling in 32-bit mode. Or possibly Gentoo optimizations. I can't see how it would make a 50%+ difference though. If so, then the V40z will probably compile much, much faster under the same conditions because there is no way your single 246 is that close to a 4-way 850 machine. No way.

Re:2.4/2.6 compile times compared -- v/s whitebox

[unixwin]

Can you post gcc -v , make -version and relevant section of Makefile? HOSTCC = gcc HOSTCXX = g++ HOSTCFLAGS = -Wall -Wstrict-prototypes -O2 -fomit-frame-pointer HOSTCXXFLAGS = -O2

Re:2.4/2.6 compile times compared -- v/s whitebox

[Cthefuture]

$ gcc -v
Reading specs from /usr/lib/gcc-lib/i486-linux/3.3.4/specs
Configure d with: ../src/configure -v --enable-languages=c,c++,java,f77,pascal,objc,ada, treelang --prefix=/usr --mandir=/usr/share/man --infodir=/usr/share/info --with-gxx-include-dir=/usr/include/c++/3.3 --enable-shared --with-system-zlib --enable-nls --without-included-gettext --enable-__cxa_atexit --enable-clocale=gnu --enable-debug --enable-java-gc=boehm --enable-java-awt=xlib --enable-objc-gc i486-linux
Thread model: posix
gcc version 3.3.4 (Debian 1:3.3.4-9ubuntu5)

$ make -version
GNU Make 3.80

HOSTCC = gcc
HOSTCXX = g++
HOSTCFLAGS = -Wall -Wstrict-prototypes -O2 -fomit-frame-pointer
HOSTCXXFLAGS = -O2

Re:2.4/2.6 compile times compared -- v/s whitebox

[TheOrquithVagrant]

My little cluster of two dual Athlon's (one machine with 2x MP 1600+, the other 2x MP 2800+, each with 512MB RAM) compiles 2.6.4 in 1 min, 57 secs. Needless to say, I'm rather unimpressed with the benchmark of this 4-way Opteron. I seem to remember one of the big 12-way IBM Power4 servers doing a kernel compile in 6 secs... now THAT is impressive. :)

Re:2.4/2.6 compile times compared -- v/s whitebox

[unixwin]

2 things are different.
Your running 32 bit 3.34 gcc
while i'm running a 64 bit, 3.2.3 version.
Hmm very interesting timings.

thx.

gcc -v
Reading specs from /usr/lib/gcc-lib/x86_64-redhat-linux/3.2.3/specs
Configured with: ../configure --prefix=/usr --mandir=/usr/share/man --infodir=/usr/share/info --enable-shared --enable-threads=posix --disable-checking --with-system-zlib --enable-__cxa_atexit --host=x86_64-redhat-linux
Thread model: posix
gcc version 3.2.3 20030502 (Red Hat Linux 3.2.3-42)

Re:I smell ...

[otis wildflower]

Hell, what server admin runs Linux, the lowest common denominator of Unix and Unix-workalies, on a real server?

Take a trip to NYC, walk out of the Wall St. 4/5 station, pick a tall building, go up on the roof, unzip your fly, and take a piss. Inside the building you hit you will find a company that transacts hundreds of thousands of dollars of business per MINUTE.

On Linux.

Better be quick though, as there's TONS of jobs moving across the Hudson :/

Just been wondering about this

[ShatteredDream]

If Apple continues to make MacOS X Server more and more robust, and if they could reduce the price on the XServes, then for many environments why not run MacOS X? From looking through guides to OSX Server, it seems really straight forward to setup and maintain compared to even most Linux distributions and looks like it just might be something that if marketed correctly could at least clobber Windows Server for many small business server needs.

I remember taking a networking class a year and a half ago where we did Red Hat 9 and Windows 2000. Even though I already was comfortable with Linux, it just seemed to be a lot easier to configure than Windows. In fact, I was actually quite amazed at how much harder it was to get Windows to do something server-related through all of the GUIs than it was to do it on Linux. Combine the fact that OSX is a UNIX clone at its core and that it's GUI is well-designed and terribly slick, I just can't imagine why most companies don't even look at it. If kept safely behind a good firewall it should be easy as hell for non-geeks to keep running for basic things like file/printer sharing.

Re:Just been wondering about this

[byronmiller]

OSX doesn't have the capacity yet to make use of "big iron" (ie large memory systems). Sun has a decade + of experience in midrange computing.

Plus the price for this sun box outdoes the price i imagine we will ever see from the likes of apple

Re:Just been wondering about this

[prockcore]

Apple continues to make MacOS X Server more and more robust, and if they could reduce the price on the XServes, then for many environments why not run MacOS X?

Because the Darwin kernel has pisspoor I/O. Which makes for a slow server.

Re:Just been wondering about this

[martian265]

Ahh, I wish I had mod points. The gp needs to be modded down, it meets the criteria for a troll or flamebait comment. The parent is spot on. Comparing an Apple XServe to a Sun AnyServer is truly like comparing apples to oranges.

The gp reads more like a form letter than an original thought.

Just been wondering about this...Is there a pro-Apple forum out there where the posters use their pretty little shavers, I mean Macs, to compose random form letters to post onto Slashdot articles that have no relevance to the topic.

While I am sure that those XServe's look much prettier next to a nice floral arrangement or a modernistic "authentic" Indian Sandpainting (made in China), they are not intended, designed or in any way capable of performing on par to real midrange servers.

Re:Just been wondering about this

[sysadmn]

Look at the target environment. There are a lot of large shops that are already running Solaris/SPARC. Choosing Sun for X86 Solaris or even Linux is easy; choosing Apple is less so. We have hundreds of Suns, from V100 to E15K; v20z and v40z fit seamlessly into our environment. If you're a small shop, with few existing servers, maybe Apple has a play.

Re:Just been wondering about this

[e40]

Don't forget that Mac OS X is where SunOS was in the 80's (in terms of stability and maturity). I used both operating systems from "day 1".

Re:Just been wondering about this

[jurv!s]

How old do you think OS X is? Did you disqualify its NeXTStep ancestry for any particular reason?

Re:Just been wondering about this

[jurv!s]

Got any references so that I can help shut those mac fanboys up once and for all instead of appearing like some FUD spouting hater?

Re:Just been wondering about this

[e40]

Your comment only makes it *worse*. If it's been around longer than I gave it credit for, then it should be more stable.

32-bit Mac OS X is the least stable (*nix) OS I use. Try this: browse an AMD mount point with the finder. If you do, you'll need to power cycle the machine to regain control.

64-bit Mac OS X is a long way off from anything approaching stability. Solaris has been 64-bit stable for years and years.

I used NeXT OS in the 80's. It was OK, but it still lagged BSD of the time in terms of stability.

Re:Just been wondering about this

[jurv!s]

Sorry, you picked the wrong person to front on. I happen to have been authenticating and automounting home directories off a RHEL 3 file server in an OS X lab for the last year and a half and it has been extremely stable.

I will assume that I am not being trolled and attempt a meaningful response. I know there are several methods for setting up automounts on OS X and I *don't* use amd. I use the Automount via File method described here. I find this way let's me use the same maps in both the linux and OS X labs (although I do add the -P flag on OS X to make it use a "secure" port).
I can't truly troubleshoot your amd problem, but I can make a guess that the Finder was your real problem and not OS X. The Finder is in many ways the most frustrating leftover from OS 9. Would you do me a favor and try to repeat your AMD lockup? I bet you that you can ssh into the OS X box while the Finder's giving you the SBBOD and issue a 'killall Finder'. If you can browse the amd mountpoint from the CLI, then you know it's the Finder. I very infrequently resort to power cycling any Mac running OS X. No more often than I'd do for my Linux lab. Your post reminds me of one of the professors that asked to test out a linux box from me. He came back the next day and had this proud look on his face as he told me that he'd *crashed* linux. Of course, he'd done nothing of the sort. He just didn't know how to restart the X Server! Of course, I do long for virtual consoles on OS X- $deity forbid we be without another box to ssh from to rescue the hung Mac :-(
To sum up- I wouldn't compare OS X to Solaris. Solaris has been refined many times over as a server OS while OS X is aimed at the desktop (where I feel it really trounces the Solaris desktop). You are however, using anecdotal evidence from one application to judge an entire OS unstable and I think you need to reevaluate that before you draw any other rash conclusions from it. Pardon the gruff tones at the end- it's late here and I'm too tired to soften my language for general consumption.

Re:Just been wondering about this

[e40]

Your ego can stand down. I didn't "pick" on you. I said I used AMD. You said you haven't. Who is fronting?

The information about the Mac OS X automounter (and link) is appreciated. Perhaps my problems will go away if I use it. We suspected as much, and it's good to have confirmation that someone uses it in a mixed OS environment.

Yes, it's the finder that is at fault here. My point is that Apple distributes software that doesn't work with basic Mac OS X functionality. That's not smart.

As for ssh'ing into the box, no, unfortunately. That was the first thing we tried. It is a hard lockup.

Re: Solaris vs Mac OS X. Yes, that's unfair, thought it should not be. Remember, Apple sell a product called "Mac OS X Server".

Virtual consoles are great, and would be a great addition to Mac OS X.

Re:What happens when you try to install windows on

[Megaslow]

It's on the Windows HCL

Re:Imagine..

[DA-MAN]

Big fucking deal, a 4-way 3u Opteron Server . . .

Appro has a Quad 1u unit. Makes for an awesome cluster . . .

I do not work for Appro, but I am a customer

$CFLAGS just kicked in, yo!

[RLiegh]

Re:I love the combination...

[agraupe]

No... gentoo != good server distro. I am a satisfied gentoo user on the desktop, and I run a very small server (alongside my desktop, just for light personal use). I don't use gentoo for the speed, I use it for the customizability. Portage is a great tool. Ideally, the Gentoo project would make portage a tool which can be put on top of other distros, as they do have advantages. Portage, at the moment, is more or less tied to gentoo, so gentoo is what I use. There are binaries for things like KDE, and besides, it's not like I need everything now. I can be patient unlike some people. That being said, these qualities do not make a good server. You might want something like Debian or slackware for that.

Re:What happens when you try to install windows on

[BCW2]

It blows up or becomes an instant zombie. Just like every other box Winblows is installed on.

How does this compare with HP?

[Anonymous Coward]

What about this ProLiant DL585 server from HP. It seems very comprable (4 Opterons, 8 or so PCI-X slots, configurable with lots of memory and storage, not to mention a similar price point). There are links to a few benchmarks on that page. Anyone have any experience with the DL585 or similar HP servers or know how they compare to these servers from Sun?

Re:How does this compare with HP?

[kaiwai]

Why purchase it from HP? I mean, honestly, what has HP have to offer apart from *really* over priced, over proprietary computer systems that are a bitch to install any operating system onto - including Windows!

Re:How does this compare with HP?

[kaiwai]

No.

Their inkjets block more times than I could shake a stick at. After 2 HP printers I have owned had been constantly getting clogged, I decided to ditch it and purchase an Epson, I've been a happy user since.

I repaired HP desktops and servers for a living, if there were ever a machine loaded with more weird proprietary crap, it would be HP. HPs x86 servers make Apple computers look like the perfectly example of computers based on 100% open standards.

Dell , IBM and SUN machines are a walk in the park. Standard motherboard, standard processor, standard everything. Throw something in it, and it just works, can't say the same thing for HP.

As for their services - they may claim to scream linux, but get a HP consultant into a room, and they'll start ramming Windows down your throat to the point that you end up asking for mercy.

Re:How does this compare with HP?

[ocbwilg]

Have you purchased an HP server lately? We use nothing but their Proliant line at work, and they're great. Contrary to your point, I have NEVER had any issues installing any version of Windows on one of our HP boxes. You can boot from the Smartstart disk, ask a few questions, and let it do the rest in 20 minutes. Or you can boot from the Windows CD and do it like you would on anything else. The only "propretary crap" that I've ever seen installed are their Insight Management agents, which are actually quite useful.

Pricewise, they're a tad more expensive than Dell, but Dell gets deep discounts on CPUs from Intel because they sell nothing but Intel. And Dell's servers (at least their 1U systems, which I have also used extensively) tend to crap out quite regularly. In general I find them highly competitive with most other entry level and midrange servers.

Re:Dual core?

[timeOday]

OK, I admit I didn't RTA:

Obviously, a dual core V40z - which is already dual core ready - will give Sun the only 3U, 8-way Opteron that we've heard of. Between dual core Opterons and continual improvements on the 90nm Opteron steppings, server administrators have a lot to look forward to this year.

Re:Not exactly...

[sarahemm]

Other than obvious things (logos, the word 'Sun' in firmware, colours of plastic, etc.), what Sun-specific changes have been made? Just curious :)

Re:I love the combination...

[HairyCanary]

Nah, skip Linux and get Solaris 10. Call it Slowlaris if you like, but I'm so sick of my Linux boxes crashing for no apparent reason, with no warning, and with no logs at all that suggest why. My Solaris boxes are rock solid stable, with uptime measured in the hundreds of days. Solaris and Opteron. The combination is killer.

Re:I love the combination...

[Anonymous Coward]

you gain much more than 0.1% of performance compiling a linux kernel

I would say it is actually less than 0.1%. It's just pure placebo

Mourn the Advent of the Opteron

[SoupIsGood Food]

This isn't a triumph. It's a travesty. It's a dangerous thinness, the latest symptom of a mass extinction that will take computing a long time to recover from, if it ever does.

If this trend continues, you will only have the option of running your choice of OSS Unix or Windows on x86. There is no future for any other sort of chip... now that Sun's all but given up, all that's left is IBM making chips for Apple and the biggest of the big iron. Even Palm and ARM is winking out, one licensee at a time...

If you're about to invent a better way to do computing on the desktop or in the server room, don't bother. The barriers to entry are now insurmountable.

Welcome to the new Wintel world. Linux is just a way of making the monopoly feel like something else, except to those of us who remember a time when real workstations and servers stomped the earth with fire and fury... SGI Indigos, DEC Alphastations, HP Superdomes, Siemens Pyramid, Fujitsu 64's. Now it's all gone. Just x86 forever and ever, Amen.

Those who have never used anything but PC's, and the glorified PC's that pass for workstation and servers, in their personal and professional life will never understand the frightening emptiness out there. Maybe that's just progress and commoditization for you, the dinosaurs dying off... but evolution does not guarantee progress. Neither do market pressures.

SoupIsGood Food

Re:Mourn the Advent of the Opteron

[rs79]

a time when real workstations and servers stomped the earth with fire and fury... SGI Indigos, DEC Alphastations, HP Superdomes, Siemens Pyramid, Fujitsu 64's."

Excuse me I have to change my shorts now.

"Now it's all gone. Just x86 forever and ever, Amen"

Segments are for worms.

Re:Mourn the Advent of the Opteron

[stox]

s/x86/IBM360/

Somehow, I have heard this story before.

Seriously, the cycle continues. Until the next chapter, which I hope will be soon.

Re:Mourn the Advent of the Opteron

[renoX]

> Just x86 forever and ever, Amen.

And? That's just the CPU, dammit! Who really cares about the instruction set which the CPU runs?

As for being glorified PCs, have you read the article?
Good LOM, no cable, Hot-swappable fan and power modules: it looks to me as a real server not a glorified PC!
OK, it is not a Stratus but still it is as good as previous Sparc servers from Sun..

Re:Mourn the Advent of the Opteron

[Frumious Wombat]

Warning: I administered most of those machines at one point or another.

While I am sorry to see the Alpha go (a great missed opportunity, and still a reason to find Ken Olsen and slap him silly), and possibly the PA-Risc, it's hard to mourn the passing of the perennial dog Sparc "we have a balanced architecture; all of our components are equally slow!", and the bipolar SGI (Version 7.x of the compiler toolchain, otherwise known as the bug of the month club). Itanium, despite being razzed continually by 18yr old Doom players, is a good replacement for most of those processors, but thanks to Intel/HP's inept marketing, it's going to be the Alpha of the 21st century.

I have run, at one point or another, VAX/VMS, MIPS/Ultrix, Alpha/OSF-1, MIPS/Irix (3.0 -> 6.5.x), SunOS, Solaris, Power/AIX (3 -> 4), and several versions of Linux(x86/IA64/Alpha), and the only old architectures I'm sorry to see gone are the Cray vector processor and the Alpha. I can now commit more science in less time for less money than ever before, and am frankly encouraged to see the 4-8way Opterons with their Hypertransport interconnect appearing from major vendors. They'll replace nicely, over time, the SGI Origins, and similar NUMA boxes.

A mass extinction is always depressing to watch, but the only real loss in that list is the Alpha, of which pieces are showing up in x86 anyway.

Re:Mourn the Advent of the Opteron

[SoupIsGood Food]

Warning: I administered most of those machines at one point or another.

Me, too. I miss them.

I can now commit more science in less time for less money than ever before

Well, duh. Of course you can. That's the nature of Moore's Law.

The thing of it is, the Opteron is only now outpacing old architectures long since put into maintenance mode. You're not able to commit as much science in as little time as you could with true high performance platforms.

That's the whole point in a nutshell... the market cannot sufficiently innovate with only competition between two processors on a single platform, no matter how many OS's it runs.

Now for the nitpicks:

Itanium... is a good replacement for most of those processors, but thanks to Intel/HP's inept marketing, it's going to be the Alpha of the 21st century.

Thanks to it's unworkable inception, unrealistic claims, stupid-long time-to-market, ludicrous price, massive power requirements and smelt-iron heat production, too. Intel killed the i960 and i860 with gross marketing incopetence, but at least those chips had some real benefits to offer. The Itanium barely keeps up with the Alpha and Fujitsu SPARC chips, while being far more difficult to implement in a workable system. ...The perennial dog Sparc "we have a balanced architecture; all of our components are equally slow!"

Sparc isn't the best choice for number crunching, but it does a lot of other things very well, most notably its low latency and graceful degredation under heavy loads. Very useful for server work, where chewing up and spitting bits out over IO as fast as possible is more important than raw horsepower. Never understood why it was so popular as a technical workstation, tho.

SoupIsGood Food

Re:I love the combination...

[Jason Hood]

hehe nice troll =)

They will lose

[Ars-Fartsica]

The problem is that people perceive Sparc to be dead...thats all it takes. Look at Alpha - fastest platform in its day but it had the stink of death even though a well-heeled company (more than one through acquisitions) was being it.

In any case I don't see how Sun can resurrect Sparc at this point, even if they were to bring a breakthrough performance product to market (doubtful if they haven't announced anything firm by 2005).

What is a "high end" chip anyway? You have fast chips and you have slow chips.

Re:They will lose

[SunFan]

Look at Alpha - fastest platform in its day but it had the stink of death even though a well-heeled company (more than one through acquisitions) was being it.

Even thought the quality of Sun's marketing dept. is certainly open for debate, it is clearly better than DEC's was.

What is a "high end" chip anyway?

One thing that differentiates UltraSPARC from Opteron is that UltraSPARC is designed to scale to over 1000 CPUs in a system. Opteron's sweet-spot is up to 8 CPUs. Otherwise, both CPUs have similar characteristics, such as ECC support, etc.

A lot of work can get done with 8 CPUs, but for everything else, there's UltraSPARC, POWER, and Itanium.

Re:They will lose

[afidel]

Well, a high end chip is one designed for throughput as well as number crunching on tiny data sets. In fact the biggest problem for most computational problems today is not cycles but memory bandwidth, and a Sparc system delivers memory bandwidth in spades for a large number of processors. The Sun machines are unfortunatly for Sun not needed by that many people as many classes of large jobs have had architectures designed that allow them to run on piles of comodity wintel/lintel servers. Sun realizes this and want to be the guy that supplies you with those comodity boxes as well as the big back end database server that feeds them all. Another fine example of a high end chip is the PA-RISC chip which does checksuming in every component and which runs all calculations through either two CPU's or two cores to make sure that hardware errors don't produce data errors. That's not something that tends to produce the fastest chip on a given process but there are companies willing to pay for it, which makes it a high end chip.

Re:They will lose

[codeguy007]

Yeah buddy, you don't know what you are talking about. The base design only ramps upto 8 procs. That will go upto 16 with dual core. However that's just if you are using just the built in HyperTransport system for SMP. If you use a custom system from cray or SGI, you can go more than 8 procs. Way more.

Cray working on 64 to 10,000 Opteron based systems

Re:They will lose

[Bert64]

Infact, the inbuilt MP systems for Itanium is shared bus, and doesn't even scale well upto 8 cpus, just like x86 doesn't.. All of the vendors producing large IA64 systems are using custom systems too.

Re:They will lose

[joib]

Also, can you agree that the features of processor A may be more suitable for building a 100 processor SMP/NUMA system than those of processor B or C?


No, since what sets 100 cpu systems apart is not the particular brand or instruction set of the CPU, but rather the NUMA chipset and assorted hardware used to connect them.

The reason big Suns are good is not due to some inherent superiority of the SPARC microarchitecture, but rather the kickass high-end systems design team Sun has.

Re:I smell ...

[Anonymous Coward]

Just be careful how the wind blows or you'll also take a quick shower. On a serious note, only after trying to purchase some property in JC and then moving back to NY from JC I realized what an overpriced, fucked up dump east NJ is. Especially when considering that east NJ was a land-filled swamp (Hamilton's project I think) and now has real estate taxes higher than Park Avenue's, and is controled by incestuous bureaucratic rednecks.

(just an idea, JC real estate tax is ~$2000/1000ft^2 per year based on property value from 6 years ago, when property value was 1/3 of current market value)

Re:I love the combination...

[Nutria]

I'm so sick of my Linux boxes crashing

So stop using crappy hardware!

What was so good about these dead systems?

[Ars-Fartsica]

Was it a modern open unix? You can have that on x86.

Was it high performance? x86 outperforms all of your examples on a per-CPU basis.

Was it incredible graphics? These geezers don't have access to modern gpus.

Was it rugged hardware? x86 boxes are now equipped as good or better than any of your examples.

I'm not sure what it is you got out of using these systems that represents a legit advantage.

Re:What was so good about these dead systems?

[Anonymous Coward]

except risc systems using 'real' unix have always been hardcore and reliable, no crap that evolved from the dos days, and yes i own 4 pc's most running linux, but I still see the attraction of workstation machines

Re:What was so good about these dead systems?

[igb]

With a monoculture, what happens if something
we've not yet thought of turns out to be hard
with x86? Ever wonder why in WW2 every air force
kept production lines running for at least fighters
and at least two bombers? Because if when they needed
an increment of performance the tails started
falling off, they had another gene pool to try
the same trick with (why did the UK keep making
Spitfires when the Tempest was clearly better
in every way? B17 vs B24? P47 vs P51? 109 vs
190)?

Had the RAF decided that the Spitfire was where the
action was in 1942 and shut Hawker down, they'd
never have had the aircraft they needed to deal
with the 262 and the V1. Had they decided that
the Spitfire wasn't going to deliver the performance
of what was coming through Hawker, they'd have
been shafted when the tails started falling off
Typhoons (elevator flutter: very hard to diagnose
in 1943).

Same's true of processors. Sadly.

ian

Re:What was so good about these dead systems?

[CaptainPinko]

I'm not sure what it is you got out of using these systems that represents a legit advantage.
Diversity for diversity's sake. It's interesting to have variety and see different approaches of doing things. It's nice to have sometihng that not just works, but something that a elegant top-down design. Some of us are pragmatists... but some of us are idealists to whom "purity" and "elegance" hold a very important value. True, in buisness none of these matter but there is more to life than buisness. Maybe these are great days of computing for buisness, but for the passionate tech enthusiast this is terrible. For a similar opinion on OSes see this: http://osnews.com/story.php?news_id=9802 .

Re:What was so good about these dead systems?

[SoupIsGood Food]

Was it high performance? x86 outperforms all of your examples on a per-CPU basis.

This is a recent phenomenon, and has more to do with the politics of monopoly and inept business strategy.

In their heydey, MiPS, Alpha and PA-RISC were neck-and-neck in terms of performance, because all were funded and developed by vibrant companies at the top of their game. Sun was slower, especially in the benchmarks, but had other advantages (like its unreal low-latency).

Then along came Rick Belluzzo, who set both HP and SGI on the Itanium/WindowsNT deathmarch, killing off R&D for all three of the top-tier RISC/Unix architectures... once HP bought Compaq, they destroyed the old DEC R&D machine, and the Alpha with it, mostly out of spite.

What would have happened if HP hadn't decided to burn its bridges for Itanium? What would have happened if SGI had hired a CEO who decided to keep them on the RISC/Unix track and to keep Mips rather than spin it off?

You would see a top teir of premium processors, and a second tier of processors x86 could almost compete with. The way it was in '97, before "Merced" and "NT" were going to be the future of technical computing.

Was it incredible graphics? These geezers don't have access to modern gpus.

Modern SGI workstations, while laboring under an antiquated processor, have GPU subsystems you gamerboys can only have fond wet-dreams about.

Even still, past history shows that with a viable high-performance oriented platform, high performance innovation takes place that takes a few years to filter down to the commodity platforms: SCSI, Fiberchannel, crossbar connections for subsystems, wide datapath expansion cards (DEC's 64bit PCI comes immediately to mind), and GPU subsystems like anything from SGI or HP's Visualize.

Commodity gear has caught up, only because of Moore's law. The vendors essentially gave up their cutting-edge workstation and server markets to push their commodity systems, thinking they would offer higher margins and a wider customer base.

Instead, Dell took everything, slashing margins and eroding everyone else's share of the x86 pie.

Now Sun is making the same mistakes.

Understandable, though, as their SPARC R&D has been a complete mess. The Fujitsu SPARC chips are kinda sexy, but getting long in the tooth.

Opteron is a last-gasp stopmeasure for Sun. It will probably do little except irritate their longterm Solaris/SPARC customers.

SoupIsGood Food

Re:What was so good about these dead systems?

[Bert64]

All of these processors were massively ahead of x86 at one time, and could still be if they had the same development effort pumped into them.. A lot of the die space on an x86 chip is wasted to support legacy applications so a modern chip would have an immediately advantage here.. Tho saying that, x86 is still being beaten on floating point performance by Alpha and POWER5 - even tho Alpha hasn't really seen any development for years.

Re:What was so good about these dead systems?

[Hal_Porter]

x86 outperforms all of your examples on a per-CPU basis

x86(64) rocks for single CPU, but as you look at systems with more CPUs it tends to fall behind.

http://www.aceshardware.com/SPECmine/top.jsp

I'm not saying that x86 is worse at this stuff than Power et al, just that there aren't many people making these sorts of systems.

If AMD have any sense at all, they'll concentrate on that stuff while Intel wastes time and money on the doomed Itanium - the margins must be higher than single CPU desktop stuff. It's be easy too, an AMD chipset for 32 or 64 way to prime the market.

Mind you there's no accounting for the dumbness of big companies. Last time I looked at their site, you could parody^Wsummarise it as

"OMFG we get 150FPS in [fashionable new game] at [ultra high resolution] we are teh l33t Intel only get 120FPS"

as if you can sustain a billion dollar company selling $1000+ CPUs to the richest 1% of PC gamers.

Re:What was so good about these dead systems?

[SunFan]

You make several good points, but I don't think Opteron is a bad thing for Sun. For once in history, there's a decent x86 chip out there, which Sun can offer to its low-end customers.

Also, Solaris x86 and Solaris SPARC are source-compatible with eachother. Applications should need only a recompile between systems--much easier than moving between Linux/x86 and Solaris/SPARC.

For customers who really like UNIX, Sun has set themselves up pretty good, IMO.

Re:What was so good about these dead systems?

[akuma(x86)]

I think the executives at SGI, HP etc... saw the writing on the wall.

High end CPU development is expensive and becomes even more expensive with each new process generation. Having everyone (MIPS, SPARC, PA-RISC, Alpha) develop their own architectures with little difference in performance between them was not cost effective. Capital investment in high end CPU development did not return enough dollars to recoup the costs.

x86 on the other hand, had plenty of PC revenue and profits to justify incredible investments in R&D. Intel was in a clearly superior financial position to take on ever more risk, complexity and R&D costs with respect to CPU development.

The executives probably crunched the numbers and realized that they could no longer afford to be in the game and "outsource" their CPU development to Intel.

Much in the same way that they "outsource" their DRAM development to the DRAM semiconductor companies.

It just doesn't make economic sense to fund a CPU design which takes a few billion dollars these days and several years of development --- with such low volumes as found in the server and workstation markets. Especially when that market is even further decimated by 3 or 4 competing RISC architectures.

Another nail in the RISC coffin is that the ISA choice - RISC vs. CISC, matters less and less with each process generation and with new architectural innovations like pre-decoded instruction caches, trace caches and store forwarding buffers. The fixed overhead of x86 gets dimished as transitor budgets grow with each generation.

Re:I love the combination...

[Zorilla]

"Spend your life compiling Gentoo kernels for that 0.1% performance gain"

Gentoo is not a kernel, and you gain much more than 0.1% of performance compiling a linux kernel. You stupid anonymous.

By that logic, I shouldn't crap in a toilet bowl because my toilet isn't a bowl. You'll master the English language someday - just not today.

2.5 minutes on a 4 way.. we did 2.7 on a 2 way

[Bruha]

Sorry but shaving a few seconds of that compile does not justify the premium of the sun name being on the box. We've built our own 2 way servers and they compile nearly as fast.

Re:2.5 minutes on a 4 way.. we did 2.7 on a 2 way

[Doomdark]

We've built our own 2 way servers and they compile nearly as fast.

There's obviously some added premium for the Sun logo, but do keep in mind that much of the price comes from premium components used (to improve reliability, and allow hot-swapping). And although kernel compilation gives some indication of performance, scalability of actual services being run on the system (db/app/web servers, most likely) is generally better than that of kernel compilation (which doesn't get fully parallel with make, due to linearity of things like linking). So 4-way system may scale nicely, and be able to roughly do the work of 2-way system (for db/app/web servers), but be bit less maintenance work.

The other thing, too, is that convenience of just buying something off the shelf is also worth something. I've built my own PC systems for years, but right now I'd rather just let someone else do the monkey work (assembly, smoke testing) and concentrate on my actual job.

Re:Since when...

[rs79]

Since when is the kernel compilarison speed is a benchmark factor for a server hardware."

That's what you say if you want slashdot to pay attention. I'm old fashioned and I benchmark servers with the application I intend to run.

Ram?

Yes please. You know the drill. If it has a power cord it needs more RAM.

Penguin Computing?

[Leghk]

We took a look at a piece of hardware that truly has a few competitors...

They list the HP as a competitor, and a few brands I've never heard of, but no mention of Penguin Computing's Altus series offering. Anybody have any experience with Penguin Computing in general, or their Altus series? We are looking into purchasing opteron servers but are having a difficult time weighing between vendor reputation and configuration flexibility (redundant power / scsi raid).

Re:I love the combination...

[BigBuckHunter]

Gentoo is linux, and BSD, Hurd, Solaris... Oh, and MacOSX.
BBH

Opetorns're for low end customers

[reachbach]

Sun's selling linux/opteron boxes only to low end customers. Remember, a linux box comes kick ass cheap and does not have half the features of S10. But for the serious ones, Sun still offers S10 on Sparc(heard of the 32 way Niagara?that's what you would call a beast of a server.A server for real internet workloads). The take home points:
1)Sun sells Linux too(surprise,surprise!!).
2)It does this for the low end guys
3)Sparc is still the defacto chip for any serious high end customer.
4)Sun's amd boxes will be far superior to those of IBM & HP. Why? 'cos HP & IBM don't have their own industry standard OSes, while sun has a beauty in the form of Solaris10 that will give you better value for money on your AMD64 processors.
Finally,learn to accept the truth.Call a spade a spade.S10 is simply a superioir OS to any other OS that exists on this planet today. Embrace it or be left behind. Use DTrace if you like S10 or be content with using top and gather cobwebs snuggling up to a cute penguin.

[ And the Sun never sets forever... :-) ]

Re:Odd advice

[ArbitraryConstant]

"you would install FreeBSD, which has had serious issues throughout its latest release cycle. I think you spend more time reading about these products than using them."

I'd probably run the 32-bit port of FreeBSD 4.x if I was going to run FreeBSD. Or possibly OpenBSD. Debian wouldn't be my first choice, but if I were a Linux guy it would be so I left it there as a nod to Linux people that think more or less like me.

so what?

[CAIMLAS]

That's not even marginally impressive - for a 4-way Operteron. I'd be quite unsatisfied.

I did a build on an IBM x306 about a month ago, from a fresh tarball, in 2 minutes and 28 seconds. The system was in RAID-1 setup (with SATA, of course). This model had (I think) a 3.2GHz P4.

Why is this supposed to be impressive?

Re:Curious

[Doomdark]

Well, I think they actually officially support multiple OSes, so Linux flavours (SuSE, Red Hat) are not the only things that it can run. Solaris x86 is obviously supported too, and maybe others (Win2k probably). And chances are FreeBSD would run on it too, although probably not officially supported. But getting the complete list would require reading TFA. :-)

Old And Obsolete...

[loony]

Hmm - I'm starting to wonder why people get excited about a box that has been on the market for years (its a newisys 4300 for gods sake, nothing new) and is going to be phased out shortly for a sun internal design? (well, not really sun - they bought the whole company this time than rather just oem the system)...

besides that, they run stable, fast and are all you'd expect from a 4way opteron... Just the NSV (network share volume) for the sp is pretty unusual... Also doesn't have a virtual CD, making it somewhat difficult if you can't network boot...

Opterons have excellent memory bandwidth

[yppiz]

Opteron-based systems have excellent memory bandwidth - 6.4 GB/s per processor bidirectional. I believe they do very well on the Stream benchmark and with real-world memory intensive applications.

--Pat

Re:Opterons have excellent memory bandwidth

[yppiz]

After I wrote this, I actually tried testing the Opteron bandwidth, and even a simple tuned read loop runs at 4.5GB/s, not 6.4GB/s

(Opteron 242, PC2700 RAM)

--Pat

Re:Dual core?

[codeguy007]

We have quad blade designs and may soon be able to do 8 way blades if the things can be cooled. Having an 8 way 3U is not that impressive.

Re:Since when...

[Lehk228]

far more have than the number of people who have run resource intensive servers on their home machine.

Re:I smell ...

[otis wildflower]

You have a weird, juvenile, and disgusting way of trying to make a point. Grow up.

Flattery will get you nowhere.

When you are a kernel hacker...

[Phil John]

...you worry more about how long it will take to compile.

Imagine the scenario: you've made a few changes to one of the kernel subsystems, you can now either:

• Sit back and wait 20 minutes whilst compiling before you can test (but when you're testing the whole system is 1-2% faster)
• Sit back and wait around 10 minutes for the compile

When you're compiling multiple times in a day the raw compiler performance matters more than the raw system performance.

That's also why some people never get the whole bootup time reduction idea, "But I keep my boxen up 24x7 and haven't rebooted for 3 years!".

Some of us reboot machines now and then when we change a config, the time it takes to come back up is important. Also, for a kernel hacker, do the changes you've made cause booting to hang?

Re:When you are a kernel hacker...

[zyridium]

Um, you seem to have no idea what I was saying: Maybe it isn't actually slower after all, it just runs some additional optimisations.

If you are hacking the kernel I would not be surprised if you first do some basic tests on a less optimised build (which should be just as fast!).... Although you would probably be surprised to know that some things are more likely to break in more highly optimised code.. Anyway.

Re:I love the combination...

[Alioth]

If your Linux boxes are crashing for no apparent reason with no warning, it's typically bad hardware (usually bad memory. I've had a Linux box suffer corruption in the swap partition due to a failing disk but stay running - and I moved the swap somewhere else without needing a reboot)

I just picked a Linux box at random on the LAN, and it sure looks like an uptime of hundreds of days to me:

Linux avro 2.4.18-3 #1 Thu Apr 18 07:31:07 EDT 2002 i586 unknown
11:42am up 238 days, 18:14, 1 user, load average: 0.08, 0.02, 0.01

Re:WAY WAY WAY.....

[chrome]

If you're buying these machines to run Oracle, the cost of the hardware is dwarfed by the cost of the Oracle licensing.

Most people wouldn't buy these things for anything other than an Oracle box, I think.

My company is looking at these sun boxes because of the support and nice LOM features, to build a 10g RAC system. I'm expecting it to kick the hell out of the old E4500s we have right now.

But, as I said, the licensing is killer. Its like 80% of the price of the whole system. Don't sweat the hardware price so much.

Re:I love the combination...

[Octorian]

Yeah, but when a Sun box crashes because of bad RAM, you actually get a meaningful death message. Heck, a lot of the time the message even tells you exactly which stick is bad and needs replacing!

I have a V20z and love it

[Natales]

This new series is remarkable. I recently was lucky enough to buy one via ebay. Although it is not an officially supported platform yet, I have it running VMWare ESX Server 2.5, with 7 Virtual Machines, 6 of them Gentoo Linux and one Windows 2003 Server. My overall CPU utilization is less than 10% average... I can't begin to tell how fast it runs...

What's so special?

[Guspaz]

It's a quad-opteron in a 4U chassis. I don't get it, what's so special about that?

I'd be much more impressed with a 1U quad opteron with 32GB of RAM via 16x2GB DDR400 and 1.5TB of storage via 3x500GB drives.

Oh wait. It's already been done. It's called Appro's 1142H, a 1U quad opteron server.

Re:What's so special?

[Guspaz]

Cluster node? What are you talking about? To quote the article:

"The V40z is an entry-level server geared for everything from data mining to CAE to database work."

So, it's not a cluster node. Where did you get that from?

I will admit, I meant to say 3U in my original post, not 4U. But that doesn't change my point; there is nothing special about Sun's server, and other servers can do the same thing in a third the space.

Re:What's so special?

[SunFan]

The GP rebuttal still stands: The Appro system has no redundant power supplies or lights-out management capabilities. Can they put four of the fastest Opterons in 1U and still cool them reliably? Also, the 600GB in disks is with IDE not SCSI.

These systems are just for entirely different purposes: one is a compute cluster node, the other is suitable for running a business.

Re:What's so special?

[Guspaz]

My point isn't so much to compare the Appro system to the Sun, so much as point out that 3U for a quad opteron is not "small" and by no means anything special. People seem to be somehow impressed with a quad opteron in a 3U case.

There are other servers with management capabilities; Dell's DRAC cards are an example. And there are other servers with redundant power supplies.

And to get back to the Appro, yes, it can cool four of the fastest Opterons in 1U reliable (A row of heavy duty blowers lends a fairly intense airflow from the front to the back of the case). And yes, 600GB is with IDE. If you want SCSI, the max the Appro can take is 219GB of 15,000RPM SCSI drives. Three drives. You can RAID-5 them if you want. No, it doesn't have remote management. What does remote management give you that a combination of software and remote reboot switches don't?

Besides, you can always fit three 1U servers in the space of the 3U sun server. That's going to get you more storage capacity, more power, more memory, and more redundancy than the Sun will. Of course, at a very large price.

And lastly, again, why is one a cluster node? There is not a single mention in the article about clusters or cluster nodes.

Re:What's so special?

[njcoder]

You don't seem to understand what the difference between these two servers are.

The person implied that the 1142 is a cluster node, not the v40z. It's fairly obvious, that was one of the main purposes of designing the 1142h.

As a standalone server, it's not very good. There doesn't seem to be much redundancy in the design (excpt for the blowers). It's very powerful and compact server but if something fails on it, you're out of luck. In an HPC environment this doesn't really matter. You the rest of the servers in the cluster will keep on working.

You mentioned using raid-5 on the hard drives. This is probably why the original poster didn't bother replying to you. First, the Appro server only supports software based raid. You can install a scsi controller to do raid but raid-5 requres 3 drives. The 1142h only supports 2 drives.

If you're using it in a cluster, that single pci-x slot is probably going to be outfitted with something lie an infiniband card anyway. The onboard raid controller of the v40z only supports raid 1 (mirroring) but with 7 pci-x slots, adding a controller isn't a big deal.

As far as your software based management question, the LOM setup the sun server has will work even if the server is powered off, or can't boot.

They're designed to fill 2 different roles. One is meant to be a workhorse you can rely on, the other is meant to be part of one.

Re:What's so special?

[Guspaz]

Cluster node? I would think that blade servers would be better suited to the task. Clustering is only one of several suggested uses in the 1142h's info page.

Just because it lacks dual PSUs doesn't mean that it's not a very good server or will fail often. PSUs don't normally fail that often, and the lack of dual PSUs only means that the server is not suited for environments in which 100% uptime is required. There are many circumstances in which 5 or 10 minutes to swap out a PSU with a spare is not a big deal.

I mentioned Raid 5 precicely because it requires 3 or more drives. This is because the 1142h in fact supports 4 drives, not 2. I think you are confusing hot-swap drive bays with internal drive bays.

I fail to see what a LOM module offers that a remote reboot port and software doesn't. What could you possibly want to do to a server remotely when it is turned off or crashed other than reboot it (Or turn it on)? Besides, you can add LOM capabilities to the 1142h by adding a PCI-X LOM card. Though I'd rather use it for a card with RAID-5 support.

Re:What's so special?

[Guspaz]

I'm an idiot. It only supports 2 drives. Never mind about that. RAID-1 it is then.

Re:I love the combination...

[goobenet]

I've seen one sparc workstation (an old-as-dirt sparc5 *not ultra*) stay alive for over 1000 days... the only reason it got rebooted was for building renovation. Wish i would've kept that screenshot. We actually built around it until it hit over 1000 days... 1002 days though, it got turned off... i don't think it ever turned back on though :)

Re:Why not?

[davecb]

Sun's a hardware company, why wouldn't they support the software and chipsets that can deliver good performance in the two- to four-way market? --dave (biased, you understand) c-b

Who would have thought?

[God of Lemmings]

"Who would have thought only a few years ago that Sun would be the new champion of Linux and AMD?" I don't know. I certainly don't think they really want to be Linux's champion. They have repeatedly demonstrated behavior implying that they intend to pull the carpet out from under our feet sometime in the future after they gain our trust. Look carefully at their patent offerings and license agreements before jumping into the fire.

Windows on AMD-64?

[mosel-saar-ruwer]

zerocool [from Sun's website]: Microsoft Windows 2000 (WHCL-certified) Microsoft Windows Server 2003 (WHCL-certified)

Kristopher Kubicki [from TFA at Anandtech]: The Sun Fire V40z is fully supported under Windows Server 2003 and (of course) Solaris, but our primary focus on this initial analysis of the V40z is under Linux.

Uhh, did I miss the release of Win-64 for AMD-64? Or are they talking about running W2K and W2003 in legacy 32-bit mode?

Re:Imagine..

[DA-MAN]

big fucking deal about the Aapro, doesnt even begin to compare to the sun that was reviewed.

As someone who has tried both, I beg to differ . . .

why is there no hotswappable power supplies?

Appro does have a 2u Quad coming out with hotswappable power supplies. They already have a 2u Dual with hotswappable power supplies.

pretty lame

I have been trolled. . .