| ActiveMac | Anonymous | Create a User | Reviews | News | Forums | Advertise |
|
|
|
User Controls |
|
New User |
|
Login |
|
Edit/View My Profile |
|
|
|
|
|
|
|
ActiveMac |
|
Articles |
|
Forums |
|
Links |
|
News |
|
News Search |
|
Reviews |
|
|
|
|
|
|
|
News Centers |
|
Windows/Microsoft |
|
DVD |
|
ActiveHardware |
|
Xbox |
|
MaINTosh |
|
News Search |
|
|
|
|
|
|
|
ANet Chats |
|
The Lobby |
|
Special Events Room |
|
Developer's Lounge |
|
XBox Chat |
|
|
|
|
|
|
|
FAQ's |
|
Windows 98/98 SE |
|
Windows 2000 |
|
Windows Me |
|
Windows "Whistler" XP |
|
Windows CE |
|
Internet Explorer 6 |
|
Internet Explorer 5 |
|
Xbox |
|
DirectX |
|
DVD's |
|
|
|
|
|
|
|
TopTechTips |
|
Registry Tips |
|
Windows 95/98 |
|
Windows 2000 |
|
Internet Explorer 4 |
|
Internet Explorer 5 |
|
Windows NT Tips |
|
Program Tips |
|
Easter Eggs |
|
Hardware |
|
DVD |
|
|
|
|
|
|
|
Latest Reviews |
|
Applications |
|
Microsoft Windows XP Professional |
|
Norton SystemWorks 2002 |
|
|
|
Hardware |
|
Intel Personal Audio Player 3000 |
|
Microsoft Wireless IntelliMouse Explorer |
|
|
|
|
|
|
|
Site News/Info |
|
About This Site |
|
Affiliates |
|
ANet Forums |
|
Contact Us |
|
Default Home Page |
|
Link To Us |
|
Links |
|
Member Pages |
|
Site Search |
|
Awards |
|
|
|
|
|
|
|
Credits |
 |
Windows supercomputer breaks petaflops barrier | |
|
||
| Time: 08:35 EST/13:35 GMT | News Source: *Linked Within Post* | Posted By: Chris Hedlund | ||
|
Petaflops. Sounds like something out of Dr. Seuss or "Harry Potter," huh? Of course, it just as easily could be a tech term -- and it is. A flops (it's singular, and stands for "floating point operations per second") is a measurement of computing performance. And a petaflops is 10^15 flops -- or 1 quadrillion mathematical calculations per second. | ||
|
Write Comment
Return to News |
Displaying 1 through 25 of 167 Last | Next |
The time now is
12:30:40 PM
ET.
Any comment problems? E-mail us |
| #1 By 15406 (216.191.227.68) at 11/18/2010 11:30:12 AM |
| Late to the party, as usual, but at least they finally made it. I laughed when I read Hilf shrugging off Linux beating Windows on the same hardware. If it had been the other way around, Hilf would have had it tattooed on his forehead. Too funny. I guess with these new supercomputers, IE can be owned by a driveby download in 0.00000000000000000000001 seconds instead of the standard 0.01 seconds. |
| #2 By 428624 (70.67.12.154) at 11/18/2010 2:27:29 PM |
|
The real story is quite interesting. It involves a bug and Linux running on more nodes.
"Windows was outperforming Linux at small workloads, and eventually hit 1.118 petaflops across just under 1,300 nodes, according to Matsuoka. But when a Windows run across 1,360 nodes was attempted, the Linpack software designed for the Windows run failed due to a memory initialization bug." "But it turns out a software bug prevented the Windows HPC Server run from matching Linux's speed and ability to run across more nodes. The bug was not in Windows HPC Server itself but rather in a software package Microsoft designed to run the Top 500 benchmarking test. Satoshi Matsuoka, professor at the Tokyo Institute of Technology, explained it to me today at the SC10 supercomputing conference in New Orleans, saying Linux's victory "was purely by chance." http://www.networkworld.com/community/node/68693 This post was edited by A-NON on Thursday, November 18, 2010 at 14:27. |
| #3 By 8556 (173.27.244.6) at 11/18/2010 2:39:26 PM |
|
#2: "But it turns out a software bug prevented the Windows HPC Server run from matching Linux's speed and ability to run across more nodes. The bug was not in Windows HPC Server itself but rather in a software package Microsoft designed to run the Top 500 benchmarking test."
So, the reason Linux outperformed Windows on this esoteric machine was not that the maker of Windows did anything wrong with the OS, it was because Microsoft wrote a buggy program to benchmark Windows. Sounds like "six of one, half-a-dozen of another". |
| #4 By 428624 (70.67.12.154) at 11/18/2010 3:16:00 PM |
|
"But when a Windows run across 1,360 nodes was attempted, the Linpack software designed for the Windows run failed due to a memory initialization bug.
Microsoft has since fixed the bug ..." "Matsuoka is interested in why Windows was able to outperform Linux in running smaller problems. Since the hardware was the same for both runs, it must come down to either the operating system or differences between the customized Linpack software packages." "Matsuoka says there seems to be little difference in performance. It should be noted that Microsoft has helped fund the Tokyo Institute of Technology's supercomputing programs. "I was very curious to see which one would be superior, both in terms of the [Linpack] algorithm, and the underlying operating system," Matsuoka said. "It was very surprising, because they were very similar in performance."" Since Microsoft is targetting smaller clusters for business users, this is a big win for them. This post was edited by A-NON on Thursday, November 18, 2010 at 15:16. |
| #5 By 143 (216.205.223.146) at 11/18/2010 9:49:24 PM |
|
Sounds like this machine can possibly simulate a nuclear detonation or preform other war game scenarios.
|
| #6 By 15406 (216.191.227.68) at 11/19/2010 12:19:21 PM |
|
#2: The bug was not in Windows HPC Server itself but rather in a software package Microsoft designed to run the Top 500 benchmarking test.
Their results were borked by a bug introduced by MS. Colour me surprised. Looks like MS is feeling the pain of being their own customer here. I would also question the neutrality of Prof. Matsuoka, considering that it's a safe bet that MS is funding this whole thing (otherwise, why on Earth would they go with a closed-source, relatively untested HPC solution like Windows??) "Matsuoka is interested in why Windows was able to outperform Linux in running smaller problems. Who cares? You don't use supercomputer clusters for small problems. They are for big problems. This just smacks of MS looking for anything positive to hang their hat on. Since Microsoft is targetting smaller clusters for business users, this is a big win for them. Um, OK. I don't know too many business users who need their own supercomputer clusters. And, again, it smacks of desperation for something to crow about. If the MS cluster caught fire and started to burn, MS would be talking about how green the cluster is because it reduced the need to heat the building. |
| #7 By 428624 (142.32.208.237) at 11/19/2010 1:08:13 PM |
|
Windows beat Linux up to 1300 nodes, and the bug caused a problem at 1360 nodes. The bug has been fixed.
It can be inferred that Windows would have beat Linux at 1360 too. Huge win for Microsoft. "You don't use supercomputer clusters for small problems" Yes you do. Didn't you read the article?? Because it uses a KVM, the cluster can be running Windows and Linux on different virtual machines, which means you can carve up the cluster into chunks for different users. Some might want 100 nodes, some 1300. " I don't know too many business users who need their own supercomputer clusters." Many do. Are you being deliberately obtuse, or are you normally this ignorant? Example: "WorldWinds, Inc. has installed one of the most powerful computer clusters in the New Orleans area at Slidell's Gause Boulevard Complex. WorldWinds purchased the system to run storm surge simulations for historical and hypothetical hurricanes that will be used by the Federal Emergency Management Agency (FEMA) to develop new flood zone maps." http://www.hpcwire.com/offthewire/WorldWinds-Installs-Computer-Cluster-in-New-Orleans-Area-to-Run-Storm-Simulations-107303263.html |
| #8 By 428624 (142.32.208.237) at 11/19/2010 1:11:30 PM |
|
Another example: Azure
"At the PDC2010 keynote yesterday, Pixar Animation Studio’s Chris Ford demoed a cloud-hosted prototype of their Academy Award-winning Renderman solution that in my opinion makes the best case for cloud computing I’ve seen yet. The Windows Azure-powered prototype website, albeit extremely polished, is an example of how the company could easily transform its industry-leading product limited in reach by its hunger for resources into a service that would allow anyone from small studios to even indie filmmakers could then take advantage of the powers of Renderman, even if they don’t have the resources to establish and support a rendering farm." http://www.istartedsomething.com/20101030/pixars-renderman-prototype-best-realization-cloud-computing-vision-yet/ |
| #9 By 15406 (216.191.227.68) at 11/19/2010 2:42:05 PM |
|
#7: It can be inferred that Windows would have beat Linux at 1360 too.
No such thing can be inferred. Windows CC, IIRC, was notorious for not scaling well. I guess we'll find out next year. Yes you do. Didn't you read the article?? The SeattlePI blog that this is linked to? Yes. It didn't say anything that supports your contention. I've been having eye problem lately so maybe I missed it, but I don't see anything in there about it. Many do. Are you being deliberately obtuse, or are you normally this ignorant? On a global scale, the vast majority don't. Most businesses don't need a supercomputer. They just don't. I don't know why you think they do. And why are you cherry-picking companies to prove your point? Seriously, you pick Pixar and a climate-modelling company? How many of the Fortune 500 use supercomputers? How many of the Fortune 100? How many out of the millions of business worldwide? |
| #10 By 8556 (173.27.244.6) at 11/19/2010 4:49:27 PM |
| #5: How about a nice game of chess? |
| #11 By 95132 (96.25.183.211) at 11/19/2010 9:13:37 PM |
|
"How about a nice game of chess? "
--- Later. Let's play Global Thermonuclear War. |
| #12 By 428624 (70.67.12.154) at 11/20/2010 12:25:49 PM |
|
#9 "Windows CC, IIRC, was notorious for not scaling well"
"Microsoft announced that Windows HPC Server has surpassed a petaflop of performance, something less than a dozen existing supercomputers can achieve." #9 "Most businesses don't need a supercomputer" "Microsoft puts the current HPC market at 15 million users, with 1 million accounting for 80 percent of the total capacity. There are another 55 million organizations that have a need - but no access - to HPC." "Microsoft demonstrated its scope by doing 100 billion comparisons of protein sequences in a database managed by the NCBI, doing the work in less than an hour and for a fraction of the cost - hundreds of dollars - it would have cost using traditional methods." http://www.networkcomputing.com/cloud-computing/microsoft-bringing-supercomputing-to-the-masses.php |
| #13 By 15406 (216.191.227.68) at 11/22/2010 3:49:56 PM |
| #12: What are you, Son of Parkkker? All your responses are words from other people, specifically Microsoft. MS is hardly an authoritative, objective source on themselves. |
| #14 By 428624 (70.67.12.154) at 11/22/2010 10:16:00 PM |
| I don't understand. You fabricated criticisms. I responded with facts. Why are you complaining? |
| #15 By 15406 (216.191.227.68) at 11/23/2010 8:14:12 AM |
| #14: There are a lot of things you can call Microsoft's PR statements, but 'facts' isn't generally one of them. The word 'fact' is usually reserved for things that are known to be true. MS routinely lies. Just last week MS was pooping on nasty FOSS people who wrote a driver for the Kinect, accusing them of hacking their precious hardware. Fast-forward 2 days and suddenly MS loves it and made it that way by design. Um..... yeah, sure you did. |
| #16 By 428624 (70.67.12.154) at 11/23/2010 1:13:54 PM |
| Explain please. Which facts do you dispute? I don't understand. |
|
Write Comment
Return to News |
Displaying 1 through 25 of 167 Last | Next |
The time now is
12:30:40 PM
ET.
Any comment problems? E-mail us |
