coComments related to insideHPC

Anon – are you still in the PE

2009-05-28T15:14:37.931+02:00

Anon – are you still in the PET program?

Patrice – the preservation com

2009-02-10T09:10:50.371+01:00

Patrice – the preservation community certainly has a different norm for the use of the word “archiving” than the HPC and IT communities in general, and I’ll grant you that we aren’t archiving data in the same sense that the presidential libraries archive papers of our past leaders. But our use of the word “archival” in this sense is completely within the bounds of the IT and HPC communities use of the word, and our data is indeed organized, searchable, managed, preserved, and (if you read the linked article you’ll see) completely intended to be accessed in the future. Thus the need to store it. I have 3 PBytes of the stuff in my center, one of those referenced in the article.

With all due respect, having b

2008-10-26T04:07:02.811+01:00

With all due respect, having been an onsite at one of the centers listed in the article, I can tell you that Mr. Henry has no clue about the validity of the data he is trying to figure out how to store and pay for. He is a bean counter plain and simple… always was, always will be. For him it always boils down to cost and innovation goes out the window. After looking at one user who had a terrabyte of Netscape caches archived, I realized this is write once, read and validate never.Just my $0.02

stgt – I’ve updated the link f

2009-02-27T08:26:30.806+01:00

stgt – I’ve updated the link for you.

Wawrzek – awesome! Thanks for

2008-12-22T10:10:51.981+01:00

Wawrzek – awesome! Thanks for the confirmation. If you are affiliated with TASK be sure to let me know when you guys are doing cool stuff; I don’t get much HPC news from eastern(ish) Europe.

Brian – I think that’s a great

2008-04-04T20:21:32.055+02:00

Brian – I think that’s a great example of why Cameron’s argument doesn’t make sense. I’m going to start using it in my own conversations with these sorts of people.

Alexander – awesome. We don’t

2008-12-22T15:46:14.243+01:00

Alexander – awesome. We don’t get news on it very often this far West, so I’d very much appreciate it if you sent interesting things related to the Top50, or just HPC in Russia in general, our way.

Andy – thanks for the correcti

2008-07-16T14:16:35.830+02:00

Andy – thanks for the correction. After reading the story I thought the money had been allocated, not just planned for allocation.Do projects tend to get funded once they’re on the list?

“R”: You are right, this isn’t

2009-05-11T21:09:19.122+02:00

“R”: You are right, this isn’t a high end machine. But this company is not a religious organization — it’s a public company. If they can make money on the CX1, or on lego models of the Cray 1 for that matter, then I say “hooray!” That’s more money that Cray will turn around and put into the next generation of high end designs.

I agree with the “Anonymous” c

2009-01-13T09:00:20.939+01:00

I agree with the “Anonymous” commenter (there are several) that points out the problems of scale here. Cray is a small company that builds bespoke computers for well-heeled clientele. How now do they make and support “off-the-rack” in a way that doesn’t swamp the company?Partners. Let’s stipulate that the people that run Cray would have anticipated this problem (Pete’s smart, so are the rest of them). They know that high-end Cray can’t support these systems in volume. So they partner on the manufacturing (leaving their high end teams undisturbed) and partner on support with some already global support force specializing in high volume. They can then put in some engineering on the front end, and as the product takes off look for opportunities to more tightly integrate it into their high end lines. Perhaps ultimately taking the approach that SiCortex has taken with their SC0072 system and pouring the SeaStar into the CX1 to make it a true development platform for the XT line.Only time will tell about the other commenters’ arguments regarding value and market acceptance, but there are a variety of ways that Cray could field the CX1 without breaking its own high end model.

SIR, We will like to know more

2008-11-18T16:40:51.910+01:00

SIR, We will like to know more about this supercomputer and are interested to distribute it in the Nigerian Market. Best Regards, Aloy

Yes, what Mike said. I spent

2008-10-20T20:48:41.162+02:00

Yes, what Mike said. I spent 18 months as part of a development team designing HPC C++ simulation software that had to run cross-platform. We were testing on RHEL 5 and HPC Server 2003. Everything ran happily on both with very similar performance. I normally like SJV's journalism, he calls a spade a spade. But to imply that Linux or Windows clusters are updated with patches on the fly like your average home user getting an Ubuntu or Windows Update is being pretty misleading. Nobody runs a cluster like that. Here, look, have an update halfway through a 6-day simulation run.... Yeah, right.

John - bravo. I wish we'd see

2008-10-18T00:58:40.350+02:00

John - bravo. I wish we'd see more of this type of balanced journalism. I suppose the sentiment during an election year runs toward shoving one's opinion around - as often as possible. Thanks for putting the "reality check" out there and reminding us all that there are 'at least' two sides to every story. Well done. Mike B.

Nope, this post isn’t snarky a

2008-10-09T21:40:09.018+02:00

Nope, this post isn’t snarky at all, no need to delete it. I’m very careful with the facts, and my readers know that I appreciate being called out when I’m wrong. I suspect you aren’t a regular reader of insideHPC. My regular readers have grown accustomed to that tone, and that comment was intended to be humorous (note my use of the phrase “wanton abandon.”) During this time the world was on fire with multicore frenzy, and I was poking fun.That said, I hope you will become a regular reader of insideHPC if you are interested in supercomputing and high performance computing applications. We tend toward to the ultra high end, but also spend a fair amount of time (and this is increasing) on the other end of the market as well.

sijarvis - first of all, we tr

2008-10-09T16:53:18.998+02:00

sijarvis - first of all, we try to keep things civil here, so tone it down a bit. Secondly, if you’re going to be snarky, pay closer attention to the dates. You posted your comment on Oct 9 of 2008 on an entry published September 12, 2007.Future aggressively snarky posts will be deleted.

Yeah ok that's fair enough.

2008-10-09T21:26:30.246+02:00

Yeah ok that's fair enough. I didn't look at the date of the original post. However, after a quick google search I found use of the word "octacore" from as early as feb 2005, over 2 1/2 yrs previous to the blog in question. Imho people posting such threads on the net must learn that reality comes before fantasy. Anyone can sit in front of a monitor and pretend they invented such and such. It's only when someone points their stories out that they at least have a chance to get a grip with the truth. Really this should have been done before. If you think this reply is "snarky" delete it. I personally see it as being straight forward. Perhaps something the computer community isn't ready for.

I was using the word octacore

2008-10-09T14:36:43.375+02:00

I was using the word octacore early last year searching for news on its development. I found several sites using the very same word "octacore" how you think you've come to be the first i don't know. perhaps you claim to have invented other things that are already in use? does the term "bollox" mean anything to you?

By the way, I sent this to a f

2008-10-08T19:08:26.166+02:00

By the way, I sent this to a friend yesterday and figure people here may like it as well. The following is a paper titled, "Predictions for Scientific Computing 50 Years from Now" by an Oxford numerical analyst (Trefethen). I think there's some fantastic points, and one that I wholly subscribe to is raised in prediction #4 - basically, determinism in numerical computing will go away. This is hard for many people to deal with and gets in the way of smart applications. Let's say I solve some system of equations to a tolerance of 10^-8, right? Now let's say I run my smart application again, and this time it selects a different parallel decomposition, some sums are done in a different order, etc. It finishes, but the number I get for this same exact input configuration is different from the previous result by 10^-13. Should I care? No, I don't think so - it's well below what I'm solving for. But lots of people care deeply. We need to change that. PAPER: http://www.ugrad.cs.ubc.ca/~cs303/2006W2/Talks/trefethenFuture.pdf - Brian, off his soapbox for now. :-)

I agree with this in principle

2008-10-08T18:57:25.827+02:00

I agree with this in principle, but let me call this the 'grand unified theory of HPC'. And just like its analogue in physics, I'd bet we're still a ways off from actually having it. Certainly, it should be pursued, but don't expect this level of sophistication soon. I'm a big proponent of 'smart' applications.. that is, choose the right algorithm, parallel decomposition, etc., based on what it is you're trying to do. Got an MD-type simulation? Ok, if you're running with a small number of particles, you may with a P-P method... going higher, you may want to use a P-M, higher still, some FMM method. A program that's well-written can choose this at runtime, and if it isn't clear, why not simply run a few iterations with each algorithm and then select the fastest? (Like ATLAS, but done at runtime.) And it isn't just the algorithm... with communication being so important, and the characteristics of communication performance being dependent upon system architecture and software, it's (in my opinion) worth it to have applications determine the best parallel decomposition at runtime, if possible. Eg, if I run a 3D mesh on 64 processors, do I want that to be decomposed as 4x4x4? What about 8x8x1? The former has a lower surface area to volume ratio of cells, whereas the latter has fewer communication boundaries. Which is better? Beats the heck out of me, but a good application can test this with a few simple iterations and then select the best one. All in all, software needs to be more flexible, but that takes time and effort, and to my dismay, I rarely find myself with free time with which to go nuts on good design. Until that changes, I don't see the software getting too much smarter. ;) - Brian, who always is long winded, it seems.

Craig, I’m disappointed in mys

2008-10-08T17:40:03.279+02:00

Craig, I’m disappointed in myself. I, of course, was not thinking of Shakespeare’s own words, but a PARAPHRASE of part of Sonnet 59. So not only did I get the author wrong, I didn’t even have the reference right. Oy.Thanks for the catch.

FYI... "What has been is what

2008-10-08T17:12:14.647+02:00

FYI... "What has been is what will be, and what has been done is what will be done, and there is nothing new under the sun." The Bible Ecclesiastes Chapter 1 Verse 9

The empty box with extra power

2008-09-16T22:51:22.756+02:00

The empty box with extra power supplies and a slidetray is $11K. Then you add three dual socket x86 boards and you're over $25K. On top of that, there is no interconnect designed- in to make this ever be a low latency MPP system. It's just a sad, costly pile of x86 boards. Marketing/repackaging hoax like the IBM iDataPlex. Very disappointing. True supercomputer engineers at Cray must be shooting themselves over this effluvia.

Cray has attempted to enter th

2008-09-16T21:05:42.747+02:00

Cray has attempted to enter the "small supercomputer" market many other times in the past. There are a number of customers running the Cray XD1 (Octigabay) and cluster systems sold to them by (generally former) Cray sales reps. And there were similar efforts in the pre-SGI days of Cray Research to grow the business with smaller systems. The problem has never been technical - its been how does a company set up to manufacture, market, sell, and service multi-million dollar supercomputers cost-effectively handle a system a fraction of that price? This was the problem with all these other attempts, and I expect will doom this attempt as well.

It's funny how you can joke ab

2008-09-16T20:01:51.611+02:00

It's funny how you can joke about something and then watch it come semi true. I used to joke about porting Microsoft Compute Cluster to the Cray XT system. This isn't an exact fit, but close. Windows Compute on a Cray branded mini system. Rich

Idea to power the datacenters.

2008-09-16T18:52:15.223+02:00

Idea to power the datacenters. Strap some magnets to Seymore, wrap his coffin in copper wire, and let his spinning produce the power.

So, taking out the marketing s

2008-09-16T18:50:16.789+02:00

So, taking out the marketing speak, what they are saying is that Cray now sells Windows PC's? Call it a deskside whatever you want, it's still a PC running windows. So much for the whole "The Supercomputing Company" idea. If you buy a car you can get a keychain with your car on it. Will Cray do the same? Buy a Supercomputer Cray and get a Keychain Cray for free?

Looks like the original articl

2008-08-16T18:43:32.793+02:00

Looks like the original article is no longer there, so it's still a bit unclear what Project Hybrid means :( Yahoo - Document Has Expired The requested document, `/seekingalpha/070518/35946_id.html', is no longer available.

Stephen - not a bit. I hope yo

2008-07-20T22:41:04.384+02:00

Stephen - not a bit. I hope you keep writing pure, unvarnished reactions to the tools that HPC gives you to get your job done. Everything that everyone has said on this thread is true. HPC machines do often perform terribly after commissioning, and users are often quick to blame a machine for their own problems.A key meta point here that the HPC community is ENTIRELY responsible for is that the machines are often so hard to use that it can be hard to tell where the truth lies in every particular instance. And before someone flames me, remember I’m not some rookie mouthing off - I’ve been fielding Top 20 HPC machines for 15 years for a user community of about 1,000, so I have actual experience on this.

To anon:What I’m saying is tha

2008-07-09T18:08:11.108+02:00

To anon:What I’m saying is that we should not assume that the system is broken simply because a user reports it as so. I’m not blaming the user for anything other than overstatement and oversimplification. So there are memory leaks in some users codes: why is this? Is the problem in the hardware, software, is the old code still applicable to this system? Are they linking against the same MPI libraries as on other systems? If so, are the versions the same? I seriously doubt that the problem lies in the “MACHINE” (I’m assuming you mean hardware here) but rather in the complex relationship of the software components running on said hardware. Every new cluster comes with a period of growing pains as the users become familiar with it and the educational resources grow. Does the user have a right to complain? Sure, Ranger is useless to him. However, statements like “unlike Ranger, it actually works” or “supercomputers aren’t getting better; in some cases, they’re getting worse” are patently false and constitute the sort of hyperbole that I was warning against taking seriously above. Remember that I said that careers can and are harmed by this sort of overstatement so its more than just a case of sysadmin sensitivity.Am I prejudiced by my own experiences? Definitely, but so is the grad student in question and I believe that I’m being a bit more fair to him than he was to Ranger/TACC.

Rich,I actually agree with jus

2008-07-09T15:39:43.087+02:00

Rich,I actually agree with just about everything you write. The HPC space is dense and inaccessible and I do sympathize with my users who have to learn and deal with complicated software stacks that have nothing to do with their research area. The issue of the inaccessibility of the scientific computing arena is separate, although related, to my original comment. given the tone of that comment, its probably obvious that I’ve had personal experiences with researchers who are quick to call something broken simply because they do not understand how to use it correctly. My problem lies not with the lack of understanding but rather with the propensity of some to assign blame to others when their codes do not work correctly. I’ve seen these allegations turn into assumptions that the administrative staff is not be doing their jobs correctly, something that I took very personally.One could accuse me of making my own assumptions here as I am conflating my own experiences with those of the greater HPC realm. We could also launch from here into discussions of the absurdity of the flop arms race or the need for better tools that exacerbate the situation that I describe. Let me just say that we should take the claims of a user that a system is “broken” with a chunk of salt.

Michael, a rebuttal of sorts,

2008-07-09T15:39:33.087+02:00

Michael, a rebuttal of sorts, Yes every new hybrid, exotic, bleeding edge system made will go through a rough patch as people learn to use them. This is a fact of life. But, why should a physics professor trying to study some strange state of matter need to learn to be a programming expert on different exotic machines every time there is a system refresh? He/she just wants to have their code run in a timely manner.I’m a systems analyst on large systems myself, and so this is what I do for a living, I love the new and unusual. For me the new and exotic is a real kick in the shorts. I read somewhere, correct me if I get this wrong, that the average Physics PhD student gets on average 250 hours of computer instruction during his or her education. That’s it. Out of a six to eight year program, only 250 hours.I can fully understand the frustration of a user when it comes to porting there code to a different machine. It can be a bloody nightmare. We joke here at HECToR that all of our new users are going to be the 50 year old plus crowd. We just installed a Vector Cray X2 to our 11k+ processor XT4. None of the new crowd of graduates has ever played in a Vector arena like this, so, big learning curve time. It’s going to turn off some of our users, it’s just a fact of life.Long rambling reply to try and say that I can see both sides of this fence. The new hybrid Roadrunner ROCKS. But on the other hand, I’m glad I don’t have to port code to it. (-:Rich

As an HPC sysadmin working in

2008-07-09T15:39:23.087+02:00

As an HPC sysadmin working in a university setting, I am all-too familiar with the negative opinions of outspoken graduate students. Often times, new or unfamiliar systems are panned by researchers as poor-performing or “broken” when the problem actually lies with their lack of knowledge on how to properly use the system in question. Parallel code suffers from a terrible lack of portability but this fact does not seem to sink into a certain class of users who assume that their codes should work the same across all systems/compilers/library versions. In my shop, entire platforms have been smeared as “poor performing” simply because our user base did not know how to properly use a complier suite other than GCC. Mere technical understandings may not be worth getting worked up over but, more seriously than that, I’ve seen careers damaged by these misinformed accusations.Cutting edge systems will tend to suffer from their own novelty: hardware combinations never seen before tend to produce unique hardware problems. However, the tendency of researchers to mistake their own lack of understanding for a broken system should not be underestimated.

Being involved in the developm

2008-04-03T21:06:02.553+02:00

Being involved in the development of Scali MPI Connect 5.6.1 and the benchmarking leading to this press release, allow me to comment:As to John Leidel’s analysis; I do not interpret Scali’s press release as to generalize and state that our MPI is 18% faster than HP-MPI for _all_ applications. If you do, you have my apologies. That was not the intent.The number comes from a particular sized system and describes the difference of the SPECmpiM_base2007 metric of said MPI implementations. This metric is derived from running 13 different applications. How much Scali MPI Connect excels as compared to HP-MPI over the 13 applications which constitute the SPEC MPI medium suite varies as documented on the SPEC result pages. Obviously, the MPI Connect advantage is not a constant 18% over all the applications!Further, our claim that the compiler has a 6% impact is sort of a worst case. Here we have compared Intel 10.1 compilers vs. PathScale 3.0. A more fair comparison would have been to also use the latest PathScale compiler, version 3.1. Unfortunately, we were unable to complete our benchmarking using PathScale 3.1 due to time constraints. If we had used it, the compiler difference would have been 4%.As to HPCer’s comment around OpenMPI; one can only speculate as to why no SPEC MPI2007 results have been published using OpenMPI. There are companies around which have a strategic decision around OpenMPI and at the same time being a member of SPEC’s HPG group. And be aware, SPEC MPI2007 is a showcase for system vendors in the HPC space. I take it that if OpenMPI is a differential, positive, value add, it would have been used.

NVISION 08 is not going to bre

2008-07-30T21:02:19.183+02:00

NVISION 08 is not going to break the world record hold by Dreamhack (The worlds largest computer festival (LAN) ) Dreamhack IS and are ALLWAYS going to be The Worlds Largest LAN Party ;) we are not going to be beaten by some crazy fat americans ;) Cheers!

No its not guaranteed, but I'd

2008-07-16T18:44:21.538+02:00

No its not guaranteed, but I'd expect both of those to get through. You have to read the docs in detail to see the timescales too. The 'child of HECToR' money is not for several years yet. Hartree will probably be sooner, but still over a year away.

It's also worth commenting tha

2008-07-16T12:02:42.977+02:00

It's also worth commenting that HECToR had a £52M contribution from the LFCF, but the overall project value was in excess of £100M. Thus it is reasonable to assume that the £65M quoted for the HECToR successor will be supplemented by additional funding from EPSRC and other research councils.

The money has not actually bee

2008-07-16T11:38:29.702+02:00

The money has not actually been allocated yet. The projects announced have 'simply' (even to get this far is good) been approved to be listed on the LFCF roadmap, which means they now have to go through the development and review of a full business & scientific case to get the money they have asked to be earmarked. So, no money committed to next generation HPC yet. What is good is that, for the first time, more than one HPC programme is on the list, showing the growing status of HPC as a strategic theme.

This is most definitely data a

2008-07-11T00:03:15.251+02:00

This is most definitely data archiving. The data archives currently present in many of the major shared resource centers retains all of your aforementioned qualities. Considering the age of the HPCMP, its simply not mathematically possible for the data archive to be useful to multiple generations. The program is only 15 years old.

What you are doing may be impr

2008-07-10T22:09:55.211+02:00

What you are doing may be impressive in terms of storage, but it is not "archiving." Archived information is organized, searchable, managed, preserved, and intended to be accessible over time and usable by future generations.

To all: Wow! I never thought

2008-07-09T20:11:34.054+02:00

To all: Wow! I never thought my little post aimed at my family would spark such a discussion. I should be more careful in the future when I write publicly. Lesson learned. I did write the post in a frustrated mood, but I don't think I was too unkind. I didn't name any names, nor get into specifics. Nothing I said about Ranger is untrue, as far as I know. I should have pointed out (as some of you have) that in the future Ranger will probably be a great machine for everyone, but it isn't one yet for me. The part about Roadrunner is pure scuttlebutt, and should be taken as such. I'm pretty sure that the code that was run wasn't at all optimized for the Cell, so that's why it ran so lousy. I'll readily admit I am not an expert programmer. I am a physicist first so a functional computer is primary for me. I hope I haven't offended anyone!

To anon: What I'm saying is t

2008-07-09T19:36:09.089+02:00

To anon: What I'm saying is that we should not assume that the system is broken simply because a user reports it as so. I'm not blaming the user for anything other than overstatement and oversimplification. So there are memory leaks in some users codes: why is this? Is the problem in the hardware, software, is the old code still applicable to this system? Are they linking against the same MPI libraries as on other systems? If so, are the versions the same? I seriously doubt that the problem lies in the "MACHINE" (I'm assuming you mean hardware here) but rather in the complex relationship of the software components running on said hardware. Every new cluster comes with a period of growing pains as the users become familiar with it and the educational resources grow. Does the user have a right to complain? Sure, Ranger is useless to him. However, statements like "unlike Ranger, it actually works" or "supercomputers aren’t getting better; in some cases, they’re getting worse" are patently false and constitute the sort of hyperbole that I was warning against taking seriously above. Remember that I said that careers can and are harmed by this sort of overstatement so its more than just a case of sysadmin sensitivity. Am I prejudiced by my own experiences? Definitely, but so is the grad student in question and I believe that I'm being a bit more fair to him than he was to Ranger/TACC.

The original poster points out

2008-07-09T17:06:29.635+02:00

The original poster points out "Many different people are seeing memory leaks using vastly different codes." Having talked to many ranger users at the teragrid meeting last month I would say that that is a fair statement. This points to a problem with the MACHINE. The knee jerk reaction of blaming the user does not seem to apply in this case. Perhaps the poster up there ^^^ should first look at his prejudices before blaming users. Unless 20+ users all of who's code work on other machines are all wrong.

"The average Physics PhD stude

2008-07-09T16:05:21.493+02:00

"The average Physics PhD student gets on average 250 hours of computer instruction during his or her education." I presume you are talking about total hours, not semester hours. I only had 7 semester hours (3 classes), or about 280 wall clock hours. Your stat sounds about right. However, prior to 2006, there were no Cell processors for Universities to use, much less teach a class on how to program. Is programming these machines hard? Yes. I shudder to think about having to program RoadRunner. If I had the hours on RoadRunner, I would take advantage of any books on the Cell architecture, any classes that the staff taught, &c. Just because a user says a machine is broken does not actually mean that it is broken. The user has to continually Learn how to get the most out of the latest hardware, from papers, books, staff, tools, and anything else they can get their hands on.

I've also had the pleasure of

2008-07-09T15:40:03.087+02:00

I've also had the pleasure of dealing with angry users from time to time. Unfortunately, I would have to disagree with the users that this is our [the HPC community] fault. Indeed, the latest round of HPC platforms don't bode well for the C++ 101-level programmer. Indeed, the hardware and software have become much more complex. Thus is the natural progression of improving performance. The latest round of hybrid architectures are going to require some very clever application analysts in order to utilize them efficiently. Even then, "its application specific..."

Rich, I actually agree with j

2008-07-09T13:25:50.015+02:00

Rich, I actually agree with just about everything you write. The HPC space is dense and inaccessible and I do sympathize with my users who have to learn and deal with complicated software stacks that have nothing to do with their research area. The issue of the inaccessibility of the scientific computing arena is separate, although related, to my original comment. given the tone of that comment, its probably obvious that I've had personal experiences with researchers who are quick to call something broken simply because they do not understand how to use it correctly. My problem lies not with the lack of understanding but rather with the propensity of some to assign blame to others when their codes do not work correctly. I've seen these allegations turn into assumptions that the administrative staff is not be doing their jobs correctly, something that I took very personally. One could accuse me of making my own assumptions here as I am conflating my own experiences with those of the greater HPC realm. We could also launch from here into discussions of the absurdity of the flop arms race or the need for better tools that exacerbate the situation that I describe. Let me just say that we should take the claims of a user that a system is "broken" with a chunk of salt.

Michael, a rebuttal of sorts,

2008-07-09T11:51:24.213+02:00

Michael, a rebuttal of sorts, Yes every new hybrid, exotic, bleeding edge system made will go through a rough patch as people learn to use them. This is a fact of life. But, why should a physics professor trying to study some strange state of matter need to learn to be a programming expert on different exotic machines every time there is a system refresh? He/she just wants to have their code run in a timely manner. I'm a systems analyst on large systems myself, and so this is what I do for a living, I love the new and unusual. For me the new and exotic is a real kick in the shorts. I read somewhere, correct me if I get this wrong, that the average Physics PhD student gets on average 250 hours of computer instruction during his or her education. That's it. Out of a six to eight year program, only 250 hours. I can fully understand the frustration of a user when it comes to porting there code to a different machine. It can be a bloody nightmare. We joke here at HECToR that all of our new users are going to be the 50 year old plus crowd. We just installed a Vector Cray X2 to our 11k+ processor XT4. None of the new crowd of graduates has ever played in a Vector arena like this, so, big learning curve time. It's going to turn off some of our users, it's just a fact of life. Long rambling reply to try and say that I can see both sides of this fence. The new hybrid Roadrunner ROCKS. But on the other hand, I'm glad I don't have to port code to it. (-: Rich

As an HPC sysadmin working in

2008-07-09T11:31:00.560+02:00

As an HPC sysadmin working in a university setting, I am all-too familiar with the negative opinions of outspoken graduate students. Often times, new or unfamiliar systems are panned by researchers as poor-performing or "broken" when the problem actually lies with their lack of knowledge on how to properly use the system in question. Parallel code suffers from a terrible lack of portability but this fact does not seem to sink into a certain class of users who assume that their codes should work the same across all systems/compilers/library versions. In my shop, entire platforms have been smeared as "poor performing" simply because our user base did not know how to properly use a complier suite other than GCC. Mere technical understandings may not be worth getting worked up over but, more seriously than that, I've seen careers damaged by these misinformed accusations. Cutting edge systems will tend to suffer from their own novelty: hardware combinations never seen before tend to produce unique hardware problems. However, the tendency of researchers to mistake their own lack of understanding for a broken system should not be underestimated.

Scali’s release is helpful to

2008-04-03T19:38:26.735+02:00

Scali’s release is helpful to the extent that it drives competition on performance in MPI implementations. I notice that Scali didn’t publish comparisons against OpenMPI this time. I suspect those numbers would be pretty close.It’s a good point that the MPI implementation can influence performance as much as the compiler, but mileage varies greatly depending on app, implementation, interconnect, and cost.

If shareholders really want to

2008-04-14T12:46:20.005+02:00

If shareholders really want to unlock the company’s value, they should question SGI’s corporate culture. The priorities seem to be brand loyalty first, competitive products second. Without a strong reality-based value proposition even the most loyal customers eventually decide that the grass really is greener on the other side of the fence.

An investment group that’s see

2008-04-03T20:46:15.561+02:00

An investment group that’s seen one of the stocks it holds drop 70% in the past year wants everyone to know that the stock is undervalued. Thanks for the unbiased opinion.