Slashdot Mirror
Windows Cluster Hits a Petaflop, But Linux Retains Top-5 Spot
Twice a year, Top500.org publishes a list of supercomputing benchmarks from sites around the world; the new results are in. Reader jbrodkin writes "Microsoft says a Windows-based supercomputer has broken the petaflop speed barrier, but the achievement is not being recognized by the group that tracks the world's fastest supercomputers, because the same machine was able to achieve higher speeds using Linux. The Tokyo-based Tsubame 2.0 computer, which uses both Windows and Linux, was ranked fourth in the world in the latest Top 500 supercomputers list. While the computer broke a petaflop with both operating systems, it achieved a faster score with Linux, denying Microsoft its first official petaflop ranking."
Also in Top-500 news, reader symbolset writes with word that "the Chinese Tianhe-1A system at the National Supercomputer Center in Tianjin takes the top spot with 2.57 petaflops. Although the US has long held a dominant position in the list things now seem to be shifting, with two of the top spots held by China, one by Japan, and one by the US. In the Operating System Family category Linux continues to consolidate its supercomputing near-monopoly with 91.8% of the systems — up from 91%. High Performance Computing has come a long way quickly. When the list started as a top-10 list in June of 1993 the least powerful system on the list was a Cray Y-MP C916/16526 with 16 cores driving 13.7 RMAX GFLOP/s. This is roughly the performance of a single midrange laptop today."
2.57 petaflops per second
floating point operations per second per second?
So it dual boots? press the option key or something to get into Windows and play Crysis?
Yeah, but on the downside, it means that asian chicks are going to start gaining weight and wanting to be "liberated" and stuff, so your sexy accent isn't really going to pay off.
It is interesting that there are 6 new entrants in the top 10. Even more interesting is the fact that GPGPU accelerated supercomputers are clearly outclassing classical supercomputers such as Cray. I suspect we might be seeing something like a paradigm shift, such as when people moved from custom interconnect to GbE and infiniband. Or when custom processors began to be replaced by Commercial Off The Shelf processors.
Legally obligatory sig : My opinions are my own... etc etc
Well, it says the hardware ran linux at X speed, and windows at less than X speed...
Actually the article doesn't say that. The hardware was different: the Linux configuration had more nodes than the Windows configuration. This *might* have been for some technical reason, or it might have been for some extraneous reason (e.g., they have better things to do with this beast than run benchmarks on it).
In any case, the difference between the Windows and Linux scores was for practical purposes insignificant. It was a *benchmark*, not a real computation. Even if the benchmark is pretty good, the mix of resources used by a real program won't match it exactly (e.g. an app that uses less floating point calculations but more memory allocations might see a very different result).
Microsoft's aim is not to run on research clusters, but to make inroads into businesses that have in-house Windows system administration and programming capabilities and might have use for high performance computing. If so, the linpack benchmark is probably close to irrelevant for many applications.
Post may contain irony: discontinue use if experiencing mood swings, nausea or elevated blood pressure.
Wait, what?
Have you ever paid attention to the OS trends in the Top500? All the proprietary OSes are disappearing. It used to be nearly all proprietary Unix and BSD. Now it's 91 percent Linux.
Here's a graph showing the demise of Unix in the Top500
http://www.top500.org/overtime/list/36/osfam
Linux doesn't scale? It fits in toasters and supercomputers. I think that's pretty good scaling if you ask me.
You could probably make the argument in 1991 when Linus smote the ground and came up with the kernel, but not anymore. You could probably even make that argument before kernel 2.0. But since then? Claiming that Linux doesn't scale well just makes you look like a Microsoft fanboy whistling while walking past the graveyard at best.
--
BMO
Yet it powers most of the top 500 supercomputers and can run on embedded platforms. If that's poor scalability, I want to know what's good scalability.
SSC
Well... "being able to run more nodes" is also a function of software. It's called scalability.
Being able to throw more nodes at a problem would certainly be a "feature" for HPC.
A Pirate and a Puritan look the same on a balance sheet.
- What kind of performance can an actual program achieve on Windows on that hardware?
Less than on Linux, and that's what counts in the end, isn't it?
Coupled with the fact that licenses eat into the budget a significant amount, Windows TCO is not the bargain that Microsoft would like you to believe.
--
BMO
From TFA: "I'm not sure why the tests were run on a different number of nodes".
No one anywhere, except in your imagination, said Windows wasn't *able* to run on the extra nodes.
I need one so I can recalculate my budget spreadsheet in a femtosecond. These nanosecond pauses are getting old.
On a lighter note, so, why isn't this stuff changing our lives? I remember in the late 90's I read a story about how gigaflop computing would revolutionize aeronautics, allowing the full simulation of weird new configurations of aircraft that would be quantum leaps over what we had. Er, have.
Can I answer my own question? I mean, can I answer two of my questions? No, make that three now. Anyway, my perspective is that the kinds of engineers who have the knowledge required to write this kind of software aren't software engineers. In fact, aeronautics is rife with some of the most horrifying software imaginable. Much of it being Excel macros. Seriously. I wrote some of it.
Equine Mammals Are Considerably Smaller
Actually, flops means exactly what the OP said: Floating point operations per second.
Whenever in an argument, remember this.
I do scientific high-performance computing, and there is simply no reason anyone would want to run Windows on a supercomputer.
Linux has native, simple support for compiling the most common HPC languages (C and Fortran). It is open source and extensively customizable, so it's easy to make whatever changes need to be made to optimize the OS on the compute nodes, or optimize the communication latency between nodes. Adding support for exotic filesystems (like Lustre) is simple, especially since these file systems are usually developed *for* Linux. It has a simple, robust, scriptable mechanism for transferring large amounts of data around (scp/rsync) and a simple, unified mechanism for working remotely (ssh). Linux (the whole OS) can be compiled separately from source to optimize for a particular architecture (think Gentoo).
What advantage does Windows bring to a HPC project?
They didn't run the Windows benchmark with all available nodes. I'd assume they didn't have licenses for every node but the researchers made it sound like they had some sort of nerdgasm because the machine could benchmark running Windows on par with Linux (only a 5% margin on a smaller cluster).
Not really. Yes, there is something of an arms race for the top500, but even after the top500 no longer lists a system it will almost certainly still be in use by someone for practical purposes other than benchmarking.
XML is like violence. If it doesn't solve the problem, use more.
I am impressed that Windows can actually scale to that type of hardware. However, my questions are:
- What kind of performance can an actual program achieve on Windows on that hardware?
Fairly good if the programmer is skilled at writing super computer applications.
- Are context switches from godawful slow memory allocation calls as painfully slow on this supercomputer as they are on the typical desktop?
It shouldn't matter too much since they would (mostly) avoid context switches by only running 1 copy of the software per core and half of windows is disabled in Super Computing edition.
- How badly does the ever-essential anti-malware suite drag down the supercomputer?
Shouldn't be needed since it should be extremely hard for malware to get into such a controlled environment to begin with.
There are other reasons Microsoft's idea is a bad one such as the higher licensing costs, no possible reason to want to write custom, non GUI software from scratch on Windows etc.
No one anywhere, except in your imagination, said Windows wasn't *able* to run on the extra nodes.
I figure it was because the testers couldn't afford the licensing fees.
Isn't this about hardware, not operating systems (other than the OS being able to support the hardware)?
No, it is about two operating systems on the same hardware, one of which (GNU/Linux) outperforming the other (Windows).
And isn't the hardware simply about how much money you have to throw at it?
No, it is also about the architectural choices.
Architectural choices are irrelevant if you don't have the funding to realize them.
If you can't afford the hardware to your fancy supercomputer, you can make the best possible choices in the word, but you're still not getting a supercomputer.
Admit nothing. Deny Everything. Make Counter-accusations.
- How badly does the ever-essential anti-malware suite drag down the supercomputer?
Shouldn't be needed since it should be extremely hard for malware to get into such a controlled environment to begin with.
Digital Fortress?
PlusFive Slashdot reader for Android. Can post comments.
But will it play Flash Video smoothly at full screen?
What the hell kind of toaster runs Linux? There's hardly any justification for a mass-produced toaster to have any logic more complex than a relay. If there's an actual consumer toaster out there on the market that has linux controlling it, I'd like to see it (and buy it)!
Scientists point out problems, engineers fix them
altslashdot.org: The future of slashdot.
I don't know if it has an OS, but my Black & Decker Infrawave toaster oven sometimes gets "hung". I sure wish there was an OS update for it, I really like it.
This toaster uses what look pretty much like toner fusing light bars to toast/cook. It is really fast, and gives off a nice evil glow when it is working.
The only problem with it is that you can sometimes confuse it into a very dangerous state - i.e. it can go FULL ON, and the "off" switch has no effect. In this state the door-interlock switch doesn't even work! (I can't even fathom the stupidity of that design... I can hear the switch click, so I guess there must be a hardware engineer's equivalent of the Daily WTF? to explain it. I am a software guy, and even I know that the safety interlock ought to directly cut the heater power, even if you want to let the LCD/button logic run. I guess the logic board is just being "told" the switch went off and is supposed to do something about it. Sheesh.)
Futhermore, pulling the plug from the wall (for any length of time) only turns it off -- until you plug it in again! Then FULL ON, no stopping!
Rather than returning it, I've just instructed everyone in the household a work-around I have discovered; When Stuck "on", simply select "toast level 1", even while it is running already. I assume this is the least cooking it can, do, then press "start"... it will turn off after about one "toast level 1" more's worth of time, then it is back to normal (until the next time you piss interrupt it at the wrong time, usually by opening the door while it is full-on, but near the end of a cycle.)
It is a strange world where my desktop PC is more predictable than my toaster oven. (In terms of not "crashing".)
This issue is a bit more complicated than you think.
You will probably laugh but banks and finances do not: Excel spreadsheets.
Microsoft HPC solution allows distribute it across many nodes.
Trust me: *huge* money are there (alas, not for you, not for me and not for science).
It's much cheaper for a bank to rent a supercomputer to calculate a heavy spreadsheet written by programming-challenged but money-wise CPA then to hire a money-challenged, HPC-wise guy to rewrite (and perpetually modify it on a short notice) this spreadsheet to FORTRAN.
no possible reason to want to write custom, non GUI software from scratch on Windows
That's a bit excessive! It does have some nice things going for it, including a fairly nice API that's been binary and source compatible for decades. There's end-to-end Unicode support in all APIs, a nice event logging and tracing system, a nice performance monitoring system (WMI), various asynchronous file and socket APIs, including advanced copy-less APIs that can tie TCP streams to specific CPU cores, etc...
Unlike Linux, Windows has a built-in volume snapshot system that supports application quiescing (not just cache flushing), exportable snapshots, advanced access-list support that is standard and consistent, etc...
Really, the biggest issue with Windows is that the source is closed, so if you need something special for a cluster, you're out of luck. "patch tuesday" is only an issue on networks which are not controlled, and a supercomputer would use a dedicated, isolated network.
Any other application built for windows has the same issue.
I work for a company doing modeling for insurance, and the software for catastrophe modeling (RMS, AIR, Eqecat,) all are windows only. The simulations and models take days to run for a large data set, and the software/modeling companies aren't about to switch off of windows for the software licensed from them, and there is nowhere else to go.
I'm a concientious
And you know for a fact that they hadn't added nodes to the cluster between when they ran it on windows and on linux, or that some nodes weren't down for service or other maintenance, moving, or a million other potential reasons?
I don't pretend to know why when I don't. You pretend to know why when you don't.
The point is to NOT have anybody to code it at all.
As a professional HPC programmer, using "days to run for a large data set" is absolutely meaningless to me.
Define large. Means different things to different people.
.