Slashdot Mirror
$208 Million Petascale Computer Gets Green Light
coondoggie writes "The 200,000 processor core system known as Blue Waters got the green light recently as the University of Illinois at Urbana-Champaign and its National Center for Supercomputing Applications (NCSA) said it has finalized the contract with IBM to build the world's first sustained petascale computational system.
Blue Waters is expected to deliver sustained performance of more than one petaflop on many real-world scientific and engineering applications. A petaflop equals about 1 quadrillion calculations per second. They will be coupled to more than a petabyte of memory and more than 10 petabytes of disk storage. All of that memory and storage will be globally addressable, meaning that processors will be able to share data from a single pool exceptionally quickly, researchers said. Blue Waters, is supported by a $208 million grant from the National Science Foundation and will come online in 2011."
nah, nevermind
Skynet ring a bell? Mark your calendars 2011... (yeah... so maybe I just got done watching the Terminator movies...)
does it run Vista? ... oh wait!
I'm glad they've given it a green light.
Imagine having all that computer power, and not even knowing if it was switched on!
"Be light, stinging, insolent and melancholy"
Can it figure out how to brew the 'perfect' cup of coffee?
Here's to hot beer, cold women, and Glaswegian kisses for all.
Yes, I know this is probably a very naive question, but has anyone here actually had the privilege of working on one of these things? I mean, what do they actually use this for?
I think it's awesome, but are there any concrete advancements that can be attributed to having access to all this computing power?
Just wondering...
If you can read this... 01110101 01110010 00100000 01100001 00100000 01100111 01100101 01100101 01101011
Too bad it wasn't a red light with all those geeks around... I know I'd be interested!
That sure would have somebody seeing red! (yuk, yuk, yuk)
Apparently, by 2020, personal computers will have the same processing power of the human brain (Kurzweil 2005). My personal computer has 2 cores, my friend's personal computer has 8 cores, so let's say 4 cores is an average. Cores double every, what, 18 months? In the next 12 years there's 144 months, which is 8 doublings. So what's that, 1024 cores? So this computer is, clearly, 195 times smarter than a human!
Or maybe raw processing power just isn't a good indication of how near or far the Singularity is, ya think?
How we know is more important than what we know.
Didn't Blue Gene/L have almost 500 TFlops sustained back in 2007? 1 PFlop by 2011 seems a little... slow.
Perhaps the architectural differences will allow for substantially higher real world performance, but by raw numbers it doesn't seem as if it's moved up very much.
Didn't Blue Gene/L do nearly 500 TFlops sustained in 2007? Doubling that by 2011 seems a little... slow. Perhaps the architectural difference will have more substantial benefits in real world performance, but by the given numbers alone, it seems like a disappointing upgrade.
Cool thing about the globally addressable petabyte. That way people writing really crappy code that don't bother thinking about their memory storage can just thrash away. And who cares about pipeline stalls.
I find it funny how the people who have never been formally trained with writing in a language (Mathematics, and just science in general) write the best codes while the majority of the IT people I see write the most appalling code I've ever seen. I think it has something to do with the fact that the science people don't pretend to know everything and are much more willing to learn something new while the IT people already know everything.
.
I understand PetaBytes, and Petaflops, but what is a Petascale?
Something to measure the mass of really large objects (like a small planet)?
Vista fast enough?
Oh I forgot, that would cost 200 peta-dollars,
so maybe they won't use vista.
Where are we going and why are we in a handbasket?
So what kind of framerate will it have in Quake?
Insightful? it's [at least an attemp] at humor
http://en.wikipedia.org/wiki/IBM_Roadrunner So wouldn't that make this the second?
I just saw The Measure of a Man episode on the Star Trek Labor Day marathon. Data has a speed of 60 Teraflops and 100 petabytes of storage. That used to seem large in the late 1980s. (Episode were Data goes on trial whether he is a machine or sentient.)
what their tech persons blood elf or tauren will look like?
Ave Molech Setting
Movie here (79MB .mov).
NCSA page here.
It will not run 32 bit linux, so of course, the admins in charge are going to bitch about the lack of adobe flash support.
What are we going to do tonight Brain?
And we all know that positrons are substantially more powerful than electrons. Therefore, these electronic petaflops are no measure for Data's positronic teraflops.
I'm really looking forward to see what Skynet can do for us.
...Apple used to use a Cray to design their new computers, whereas Seymoure Cray used an Apple to design his.
More compute power is nice, but only if the programs are making efficient use of it. MPI is not a particularly efficient method of message passing, and many implementations (such as MPICH) are horribly inefficient implementations. Operating systems aren't exactly well-designed for parallelism on this scale, with many benchtests putting TCP/IP-based communications ahead of shared memory on the same fripping node! TCP stacks are not exactly lightweight, and shared memory implies zero copy, so what's the problem?
Network topologies and network architectures are also far more important than raw CPU power, as that is the critical point in any high-performance computing operation. Dolphinics is quoting 2.5 microsecond latencies, Infiniband is about 8 microseconds, and frankly these are far far too slow for modern CPUs. That's before you take into account that most of the benchmarks are based on ping-pong tests (minimal stack usage, no data) and not real-world usage. I know of no network architecture that provides hardware native reliable multicast, for example, despite the fact that most problem-spaces are single-data, most networks already provide multicast, and software-based reliable multicast has existed for a long time. If you want to slash latencies, you've also got to look at hypercube or butterfly topologies, fat-tree is vulnerable to congestion and cascading failures - it also has the worst-possible number of hops to a destination of almost any network. Fat-tree is also about the only one people use.
There is a reason you're seeing Beowulf-like machines in the Top 500 - it's not because PCs are catching up to vector processors, it's because CPU count isn't the big bottleneck and superior designs will outperform merely larger designs. Even with the superior designs out there, though, I would consider them to be nowhere even remotely close to potential. They're superior only with respect to what's been there before, not with respect to where skillful and clueful engineers could take them. If these alternatives are so much better, then why is nobody using them? Firstly, most supercomputers go to the DoD and other Big Agencies, who have lots of money where their brains used to be. Secondly, nobody ever made headlines off having the world's most effective supercomputer. Thirdly, what vendor is going to supply Big Iron that will take longer to replace and won't generate the profit margins?
(Me? Cynical?)
It's a small world and it smells funny; I'd buy another if it wasn't for the money; Take back what I paid (SoM)
The amount of porn you can download with this thing? Isn't that the number one thing the computer has evolved to?
Anything and Everything about the Net
in 40 years some kid will laugh at your pathetic attempt at geek coolness when you mention the Bluewater and say "wow your old, Im amazed anyone needed a warehouse just for one petaflop even my Wango-matic game cube has 50 petaflops"
My taxes? My laundry?
Impetuous! Homeric!
Folding @ Home easly trounces this puny supercomputer.
The world's first sustained petascale system.
Sorry if I come off as Ameri-centric, but it's been a long damn time since I've heard "Biggest *** now planed" or "New big science project for..." and it turns out to be in the USA. It used to always be like that, but recently it's been all EU/Dubai/China or something. At least occasionally some things are still done big in the USA.
Nah but it will finally run Vista.
"So long and thanks for all the fish."
in 40 years some kid will laugh at your pathetic attempt at geek coolness when you mention the Bluewater and say "wow your old..."
Forty more years of the kids saying "your"? Kill me now! :)
(That was SUPPOSED to be a preview while I got the links right. Here it is with the links fixed. View them in order...)
You know, that IS impressive but... Can it figure out how to brew the 'perfect' cup of coffee?
I think you meant tea.
No, I think Xaedalus meant the 'perfect' cup of coffee.
Bantam Dominique roosters crow a four-note song. Once you've heard it as "Happy BIRTHday" you can't NOT hear it that way
You all think this is funny. But come 2011, the year that terminator is supposed to come, we'll see who's laughing then...it all starts with a supercomputer that plays chess and then all of a sudden it wants to poop and play basketball. At that point in time, I'll be laughing with my hands in my pants and my fingers on a stick.
oh never mind.
but can it run Crysis?
I don't know where TFA got the "globally addressable PB". I think someone was misquoted.
I can't find any mention of it in the NCSA webpages, and no shared memory system exists on this level, ccNUMA or otherwise (NASA Ames has a 4TB altix system, which is evidently the largest in the world that is publicly acknowledged).
Software distributed shared memory hasn't really gone anywhere either, so I think someone was fantasizing when they wrote the article... globally accessible filesystems, sure, but shared memory is something else altogether...
UBU
The people that appllied for the grant are looking forward to, "the best game of freecell ever!"
Changing the world... one research project at a time.
I'm glad they've given it a green light.
Me too. It seems like Urbana sure learned their lesson about giving big, powerful computers red lights about seven years ago.
Never shake hands with a man you meet in a fertility clinic.
While it's not always an issue to those that care about the final result, I wonder how much power this sucker is gonna drain from the local power grid. I can see it now.. A prof making a call to the local power company, telling them, "Please be advised, we're gonna black out Urbana-Champaign for the next couple of minutes." Too bad they decommissioned that research nuclear reactor they had back in the '98. While it never was used for electricity, it probably could have been fitted. Then again, with all the wind they're used to, maybe a couple of wind turbines would be sufficient.
But only on minimal settings.
At least I learned it wasn't the term used to describe the Windows Vista marketing campaign...
And this is where AI comes in. If I could make a suggestion, the computer would be used by programmers who are really lazy, not just lazy enough to solve repetitive tasks, but lazy enough to write programs that would write programs to solve repetitive tasks. Take your average adept programmer who says, "I am too lazy to grind coffee beans, pour water and all that jazz, I'll write a program that will manage the process for me." That programmer is lazy, but a really lazy programmer is the one who studies human interaction, available components and programming languages then writes a bot to spread itself to send spam about a Russian company offering a two million dollar prize to be awarded to the best published coffee making program and then gets his new IRC buddy to figure out the logistics of spreading it for him.
Okay look, people are expensive right? I mean, buying one is expensive and all, even developing one when the starter kit is free, but renting one, and renting a talented one at that? So if you're gonna rent *hire* a really expensive human to work on a really expensive computer, you don't want him doing a bunch of grunt work, particularly when said really expensive machine is designed to do grunt work really (really, really) fast.
So what you do is get your very expensive programmer to write code for your very expensive machine that will do the work which you could have hired less expensive programmers to write. The programmer writes applications that test themselves, try variations and compare the results. Then you feed the system lots of programming tools and problems and let it sort out for you how it can best utilize its resources. Of course lots of stuff doesn't work well, maybe even most of it, but there are relatively small numbers of ways that most things can be combined, say a couple hundred million or so, and the very expensive machine can whip out those combinations, trials and comparisons while you grab a cup of coffee. After a lifetime of computing, literally minutes on such a monster, it is ready to throw massive experience and streamlined code combinations at your particular brand of personality problems, perhaps even calculating them to six significant digits for you. We're talking Eddie here, not Deep Thought, but perhaps Eddie can get you started on Deep Thought if you owe your psychiatrist that much.
Oh, and yeah, the truly lazy programmer is the one that writes up the paper that gets people with the money to sponsor such a project. He probably works at Standford University and was that guy now only vaguely remembered by Larry Page and Sergey Brin who giggles into his coffee everytime somebody says "Google it."
B) Eliminate all the stupid users. This is frowned upon by society.
I was once employed in a position where I created detailed performance/reliability models for large supercomputers BlueGene/L, etc.
Say you have an application that is infinitely parallelizable [over idealistic assumption]. Adding processors (and ignoring the communications overhead, etc) speeds up the application -- only up to a point.
At some point, adding processors starts to slowdown the entire application. Why? The probability diminishies that all processors will be up for long enough for the application to finish. Even if spare processors are available and the distributed application uses checkpoints, this effect still occurs.
Say a single node/processor has a mean-time-to-failure (MTTF) of 5 years (157680000 seconds). Two hundred thousand nodes have a MTTF of *approximately* 788.4 seconds (it's actually worse). In other words, there is probability of (1/e) [roughly a third] that 788.4 seconds will elapse without any failures. Wouldn't it just be cheaper & easier to have a 20k node computer and an application that runs for 1hr instead of 10 minutes on 200k nodes?
Yes, you could use 3/4 of the processors for active computation and have the other 1/4 as hot spares/etc... But wouldn't it just be simpler to use fewer processors in the first place? I'm not even convinced there are applications that can be efficiently parallelized over 50k nodes, much less than 200k nodes. When communication overhead and redundancies are taken into account, the utility of much more than a few thousand nodes starts to drop radically.
I've also noticed that those in the "supercomputer" field tend to have Computer Science or Physics backgrounds. These developers are more focused on obtaining exact results, which leads to very slow applications. I suspect there are very accurate (and fast) approximations for many the calculations in their applications. They use distributed application frameworks (MPI) that are fairly low-level and rigid. This means complex applications that run (slowly but well) on 1k nodes may not even be scalable to 100k nodes.
In short, 200k nodes cannot be used efficiently for any meaningful amount of time. For long running applications (a few hours), there is little need to use more than a few thousand nodes.
Aside: Don't intelligent people have anything else better to do than to blow each other up?
Sweet! A million petahertz for only $208!
Well you don't /expect/ Aero with that do you? Sheesh... You'll need over 9000!!1! video cards for that.
"So long and thanks for all the fish."
You guys realize these supercomputers purchased by the government sit idle and retire early. These things are essentially corporate welfare for the politically connected. IBM can't make a profit selling computers in the open market so they grease a few politicians and bureaucrats to get contracts to make new clusters. Jack up the price and sell support services. It's a racket.
A million petaflop computer for $208? That's amazing!!
Simulating nuclear explosions.
Deleted
Got a citation for that? Because I call bullshit. I know that you Americans have a fair percentage of the top500 but I severely doubt that a machine like this would sit idle. It would be attracting all sorts from out of the wood works. It may take a while to get to a sustained load greater than 50% (takes a while for people to tune codes) but then it'll start to fly.
.
Can I please get the computing power in terms of the human brain?
Blue Waters will =not= be the first sustained petaflop machine. That title goes to RoadRunner. Blue Waters will likely not even be the second or even third sustained petaflop machine. By the time it is actually deployed in 2011, machines approaching 10 PF will be rolling out.
I guess blue LEDs are a thing of the past for extreme gaming conputers.
It will not run 32 bit linux, so of course, the admins in charge are going to bitch about the lack of adobe flash support.
They can run adobe flash enabled apps in a virtual machine (shrugs)
With that monster, you could do one of the following: 1) Fire up 34-35 Mongrel instances 2) Launch NetBeans in under 4 minutes 3) Get 45 FPS on the latest FPS 4) Balance Bill Gates' checkbook BTW, at $45B, Gates could buy over 200 of these...and then network them.
Remember, licking doorknobs is illegal on other planets.
I'm glad they've given it a green light. Imagine having all that computer power, and not even knowing if it was switched on!
Oh, I think all the lights in the building getting dim would be sufficient.
Why, without your clothes, you're naked, Miss Dudley!
The real reason you would need this much power is to figure out how much you owe the IRS.
Isn't folding at home well into the Petaflop territory now? http://fah-web.stanford.edu/cgi-bin/main.py?qtype=osstats
Think Deeply.
Actually, I was referring obliquely to a short story by Charles Stross where a bunch of java-heads destroy the earth in the pursuit of brewing the perfect cup of coffee. But I like your interpretation as well.
Here's to hot beer, cold women, and Glaswegian kisses for all.