Petaflops? DARPA Seeks Quintillion-Flop Computers

coondoggie writes "Not known for taking the demure route, researchers at DARPA this week announced a program aimed at building computers that exceed current peta-scale computers to achieve the mind-altering speed of one quintillion (1,000,000,000,000,000,000) calculations per second. Dubbed extreme scale computing, such machines are needed, DARPA says, to 'meet the relentlessly increasing demands for greater performance, higher energy efficiency, ease of programmability, system dependability, and security.'"

Comments

Make sense, dammit

[Lord Grey]

From TFA, written by Michael Cooney and propagated by the summary:

Dubbed extreme scale computing, such machines are needed DARPA says to "meet the relentlessly increasing demands for greater performance, higher energy efficiency, ease of programmability, system dependability and security."

It looks like these "extreme scale computing" systems are needed before things like "ease of programmability" can be acheived. I call bullshit.

The actual notice from DARPA is named Omnipresent High Performance Computing (OHPC). From the first paragraph of that page:

... To meet the relentlessly increasing demands for greater performance, higher energy efficiency, ease of programmability, system dependability, and security, revolutionary new research, development, and design will be essential to enable new generations of advanced DoD computing system capabilities and new classes of computer applications. Current evolutionary approaches to progress in computer designs are inadequate. ...

That makes a lot more sense.

Now, will someone please go and smack Michael Cooney up the back of head for writing like that?

Re:Make sense, dammit

[Animats]

Right. If you actually read the announcement, it's not that they want yet more boondoggle supercomputing centers. What they want is more crunch power in small boxes. Read the actual announcement (PDF). See page 17. What they want is 1 petaflop (peak) in one rack, including cooling gear. The rack gets to draw up to 57 kilowatts (!).

Re:Make sense, dammit

[Anonymous Coward]

Quick napkin math:

Rack has 42U

SuperMicro servers (TwinX) have 2 "blades" per 1U rail slot.

Each blade has 2 6-core Intel Nehalem CPUs generating approximately 225 GFLOPS each, or 450 per U.

18.9 TFLOPS per rack and consuming a peak of over 78,000 BTU and 600 amps and 72KW (breaking the budget).

Yep, there's a long way to go. Guessing some sort of customizable GPU massively parallel system. It'll be a bitch to develop for, but probably what's required to reach these numbers.

Re:Make sense, dammit

[evilbessie]

See the moon there, DARPA want that on a stick, with sugar and a cherry on top if you please.

how sweet and innocent of them!

[zero.kalvin]

Call me tinfoil hat wearer, but me thinks they want a faster way of cracking encryption...

Re:how sweet and innocent of them!

[Entropius]

Good luck. I can encrypt something in polynomial time (quadratic, isn't it?) that it takes you exponential time to encrypt.

Re:how sweet and innocent of them!

[SirGarlon]

Actually, the military being able to crack encryption is in some sense a Good Thing. It enables them to conduct espionage and counter-espionage against adversaries such as North Korea and Al-Quaeda. Yeah that's kind of a Cold War mentality, but what is "cyber warfare" if not Cold War II?

Re:how sweet and innocent of them!

[Yetihehe]

but what is "cyber warfare" if not Cold War 2.0?

FTFY

Re:how sweet and innocent of them!

[maxwell demon]

Sorry, but DOS attacks are utterly outdated. Today you use Windows for your attacks.

SCNR

Re:how sweet and innocent of them!

[John Hasler]

But you'll have to fully deploy your longer keys long enough before they deploy their exaflop cracker that none of the inadequately-protected messages already in their possession are useful to them.

I suspect that simulations are more interesting to them, though. Think what they'd save on testing if they could fully simulate hypersonic flight and scramjet engines (not that I don't think they'll use this for cracking).

Re:how sweet and innocent of them!

[SirGarlon]

Since when the espionage is a GOOD thing!!!!!

Since September 11, 2001.

Or you could go back further, to July 26, 1939. But the real answer is, espionage has been a good thing ever since there have been enemies.

I for one am all in favor of having fewer enemies. But for the ones that can't be ignored or reconciled, espionage is a Good Thing.

Re:how sweet and innocent of them!

[$RANDOMLUSER]

If I'm North Korea or Al-Qaeda or "Red" China, or any one of a million other defined-as "bad guys", I'm not using RSA or some such, I'm using one-time-pads or steganography on any one of a billion different chat boards, probably one where I can post JPEGs. Places where the message location and encryption itself is all the sender signature it needs. It's the bankers and the private citizens (and possibly some foreign diplomatic services) who are using RSA and public-key type ciphers that (might maybe potentially could be) cracked by lots and lots of computing power.

Meanwhile, this is perfect paranoia-food for the "ECHELON is reading my e-mails and SMS!" types. Thing is, they're probably right.

Exaflops

[maxwell demon]

Quintillion is not an SI prefix. The next step after Peta is Exa.

Re:Exaflops

[daveime]

Nope, Quintillion is a quantity, whereas Petaflops, Exaflops etc are rates of calculations per second. Please don't mix your units in your haste to appear smart.

Re:Exaflops

[godrik]

for the record, there was a bunch of talks in IPDPS 2010 ( http://ipdps.org/ipdps2010/2010_advance_program.html ) about build exaflop mahcines including a keynotes.

Re:Exaflops

[Chowderbags]

FLOPS is not an SI unit.

Re:Exaflops

[Pojut]

Just like people complaining how in Star Wars, Han Solo said he made the Kessel Run in less than 12 parsecs...yes, we know that parsecs are a measure of distance, Solo was talking about being able to complete the race using a shorter route than the standard 18 parsecs, which is why a measure of distance makes sense.

Source.

Disclaimer: some people may shout "retcon" at this explanation, but at this point singling out each instance retconning in the Star Wars universe is a wasted effort.

Re:Exaflops

[hoggoth]

Stop using antiquated units! The current unit in fashion is Vuvuzelaflops.

Re:Exaflops

[DragonWriter]

FLOPS is not an SI unit.

True, that: FLOPS communicates a combination of the SI unit (1/s = Hz) with the identity of the thing being measured (floating point operations). It's like if you had KOM as an abbreviation for kilograms of milk.

Re:Exaflops

[amchugh]

Quintillion is a different quantity in long scale countries (10^30) vs short-scale countries (10^18), which is partly why the SI units were standardized.

Re:Exaflops

[Pharmboy]

A metric assload is roughly equivalent to 2.2 Imperial assloads. Hope that helps.

Peta-flops

[Anonymous Coward]

I'm glad DARPA is finally making a move to make their computing more animal friendly.

Translation

[Rik Sweeney]

I want to run Crysis 2 in software rendering mode

I Love DARPA

[sonicmerlin]

They come up with ideas that only ultra-geeks and science fiction nerds could come up with, and then they get billions in funding for it! It's like paradise. The fact that they're actually successful at advancing human technology is just icing on the cake.

Re:I Love DARPA

[MozeeToby]

Most people don't realize it but DARPA can best be described as a few dozen scientists and engineers with large checkbooks and a big travel budget. They go around the country and around the world looking for technologies that are beyond what we can do today but might be possible with the right funding in the right places. Most importantly, they're aware that a large percentage of the projects that they fund will end in failure (or rather, will not meet all their goals), but the benefits of the ones that don't outweigh the costs.

Re:I Love DARPA

[Courageous]

It's even more interesting than that. If DARPA begins succeeding a lot, DARPA seniors end up having to explain to congress (yes, directly to congress) why it is they aren't forward-leaning enough. I.e., DARPA programs are expected to fail often, and congress uses this failure rate as pro forma information about how "researchy" DARPA is.

Joe.

For security?

[LinuxInDallas]

Norton bogs my computer down too but that is just crazy :)

What's the need?

[Just Some Guy]

First, I'm entirely ignorant of supercomputing. I don't know the first thing about it. I'm asking this out of sheer lack of knowledge in the field:

What do you need a computer that fast for?

I mean, specifically, what can you do on something that fast that you couldn't do on one 1,000 (or 1,000,000) times slower? What kind of tasks need that much processing power? For example, you normally hear about them being used for things like weather simulation. Well, what is it about weather simulation that requires so much work?

The whole idea is fascinating to me, but without ever having even been near the field, I can't imagine what a dataset or algorithm would look like that would take so much power to chew through.

Re:What's the need?

[Yoozer]

What do you need a computer that fast for?

Simulating exploding hydrogen bombs, weather simulation, brute-force cracking, etc. Basically any distributed project you can think of (see BOINC) can also be done with a supercomputer.

Well, what is it about weather simulation that requires so much work?

It's a scientific model with a boatload of variables and dependencies. Ask these guys.

Re:What's the need?

[Hijacked Public]

Well, what is it about weather simulation that requires so much work?

The enormous number of variables, mostly. Weather, nuclear bombs, ocean currents, cryptography, even things as seemingly simple as modeling air flow around an object. If you are looking to develop a model of a process that involves a few thousand variables and you need to know the interaction of those variables several levels deep....you need to make a lot of calculations.

It hasn't been all that long that computers have had the computational power to dominate humans in games as 'simple' as chess.

Re:What's the need?

[Chowderbags]

I mean, specifically, what can you do on something that fast that you couldn't do on one 1,000 (or 1,000,000) times slower? What kind of tasks need that much processing power? For example, you normally hear about them being used for things like weather simulation. Well, what is it about weather simulation that requires so much work?

Theoretically there's nothing you can't do on a supercomputer that you couldn't do with an ordinary desktop computer (except possibly for memory constraints), but for that matter you could also do everything by hand. The thing is, when your problem space is very large (i.e. calculating all interactions between X number of objects, where X is some huge number, or solving something like the Traveling Salesman Problem), you are limited in your options of what you can do to get results faster. If you're lucky, you can find some speedup of your problem (I.E. going to a better level of O-complexity [O(2^N)->O(n^2) would be a huge speedup, but doesn't happen often]), or tossing more resources at it. Yes, it'll still be slow, but if it takes you a year to do on a supercomputer, that's quite a bit better than spending 1000 years waiting on a regular computer.

Re:What's the need?

[John Hasler]

I mean, specifically, what can you do on something that fast that you couldn't do on one 1,000 (or 1,000,000) times slower? What kind of tasks need that much processing power?

Detailed, 3-D simulation of things like nuclear explosions and scramjet engines.

For example, you normally hear about them being used for things like > weather simulation. Well, what is it about weather simulation that requires > so much work?

Accuracy. Weather Prediction

Re:What's the need?

[maxwell demon]

Imagine a simulation in 3D space. You model the space by a cube of 100x100x100 grid points. That's one million data points. Now say you have to do some calculation on them which scales quadratic in the number of data points. Say you manage to finish the calculation in one hour on some computer.

OK, but now you notice that those 100 data points in each direction are to inaccurate. You need 1000 points to be reasonably accurate. So now your data set is not one million, but one billion data points. And your O(N^2) algorithm makes sure that this factor 1000 in the number of grid points ends up as a factor one million in your computing time. So now the calculation would, on the same computer, need one million hours, or about 114 years. You almost certainly don't want to wait 114 years to get your results.

Re:What's the need?

[Chris Burke]

There are broad classes of algorithms where you can make good use of essentially arbitrary amounts of computing power to get better answers. When doing physical simulations of something like airflow over a jet wing, or the movement of a weather system, or the explosion of a hydrogen bomb, you'll break everything up into tiny units that you treat as monolithic elements whose behavior can be treated relatively simply, and calculate what happens to them over some tiny timescale, call the result the new state of the universe, and repeat. This is called "finite element analysis".

Because you're calculating everything in discreet steps, though, errors creep in and accumulate. The more processing power you have, the more elements you can use and the smaller time scales you can calculate over and get a more accurate answer in the same amount of time. The reason it's unacceptable to do the same calculation but have it go 1,000 or 1,000,000 times slower is that these simulations might already take hours, days, weeks, or even longer. Even the longest DoD contract needs an answer to the behavior of a proposed jet fighter wing in less than 1,000,000 days. :)

Scientific computing is an area where there will always be a use for more processing power.

There are other areas where it can be important, when you have real time constraints and can't just reduce your accuracy to make it work. I recall a story from advanced algorithms class where a bank was handling so many transactions per day that the time it took to process them all was more than 24 hours. Obviously this was a problem. The solution in that case was to modify the algorithm, but that's not always possible, and you need more computing. This is a little different in that you need the extra power to allow growth, as opposed to science where you could hand them an exaflop computer today and they'd be able to use it to its fullest.

Re:What's the need?

[koxkoxkox]

If you take weather simulation :

At a given point, you have a bunch of physical equations taking a set of parameters at time t and giving you these same parameters at time t+1. Of course, the smaller the time step, the better the result.

To have the best possible result, you should consider the whole globe at once (think phenomenon like thermohaline circulation for example). However, you should also consider the finest grid possible, to take into account the heterogeneity of the geography, the local variations due to rivers, etc. It is also important to consider a three-dimensional model if you want to transcribe the atmospheric circulation, the evaporation, etc.

I forgot the exact numbers, but Wikipedia gives an example of a current global climate models using a grid of 500,000 points (see http://en.wikipedia.org/wiki/Global_climate_model ), which is a pretty coarse resolution, working with tiles of tens of thousands kilometer square.

With the current computing capabilities, we can not go much farther for a global model. This is already an impressive improvement compared the first models, which were two dimensional and used very simplified equations, overlooking a large number of important physical mechanism.

At the same time, we have satellite data several orders of magnitude more precise. Data from the satellite ASTER were computed to provide a complete altitude mapping of the globe with a theoretical resolution of 90 m. The vegetation cover can be obtained at a resolution of 8m using commercial satellite like FORMOSAT-2. Even the soil moisture can be measured at a resolution of around 50 km thanks to the new satellite SMOS.

These sets of data are already used at the local level, for example to model the transfer between the soil and the atmosphere, taking into account the vegetation (SVAT modelling). It makes no doubt that a global climate model using a more precise grid and these data would significantly improve its prediction.

Re:What's the need?

[chichilalescu]

In fluid dynamics simulations (which include weather stuff), there are huge computational problems. I work in the field, so bear with me a little.

The best model we have so far for fluids is to use balance equations (look up the Navier Stokes equations). This means that in order to describe the evolution of a fluid in a given domain, we need to split the domain into small cells, and then integrate numerically the balance equations. To put it simply, you have to integrate numerically a system of ordinary differential equations with many many variables (degrees of freedom).
For a simple but "correct" Navier Stokes simulation, the number of degrees of freedom is proportional to Re^(9/4), where Re is the Reynolds number (the memory requirements are proportional to the number of degrees of freedom). This Reynolds number, for typical systems (like the atmosphere) is of the order of at least 10^4-10^6 (you can look up typical values on wikipedia if you're interested). Furthermore, the number of timesteps needed for a "correct" simulation is proportional to Re^(3/4).

But these are not the most complicated simulations that are to be run on such machines. Research for issues like controled nuclear fusion needs to address much more demanding problems.

Numerical simulations of physical systems are inherently hard, because they scale polynomially with their complexity. However, they are generally cheaper than actual experiments, and you have access to more data.

Re:What's the need?

[frank_adrian314159]

You almost certainly don't want to wait 114 years to get your results.

You know, back in the day, we had some patience. Plus, the notion that one would have to wait 114 years to get results made us develop better algorithms, not just throw cycles at a problem. Kids these days... Now get off my lawn!

Re:What's the need?

[Orp]

Actually, there are only a handful of variables in a weather simulation. For a typical cloud-scale simulation you have the three components of wind, moisture, temperature, pressure, and precipitation variables. Say, 13 variables. That is not why you need supercomputers.

The reason you need supercomputers to do weather simulations is all about resolution, both spatial and temporal. Weather simulations break the atmosphere into cubes, and the more cubes you have, the better you resolve the flow. All weather simulations are underresolved; to properly model the turbulent flow in the atmosphere you need to get down to cubes that are roughly a centimeter on a side. As you double the resolution (halve the length of each of the four lines that makes up a cube face) you require eight times as many cubes. In weatherspeak, we talk about gridpoints instead of cubes where it's understood that each gridpoint represents the center of one of these cubes. In the computer model, they are represented as three dimensional floating (or double precision) point arrays. So take a 3D array and double the number of calculations on each of the thee for: loops, and you've got eight times as many calculations and eight times more memory required.

And it gets worse. When you double the resolutions, you need to halve the time step. Weather models step forward in time in discrete intervals, and now in addition to more calculations for each time step (eight times as many for doubling the resolution in three dimensions) now you need to go in steps that are half as large. This means 16 times more calculations, and eight times as much memory, to double the resolution.

And many of the calculations that are being made in the innermost loop involve things like divides, non-integers powers, square roots, etc... expensive calculations. And then because it's a massively parallel simulation, you have to do internode communications - which adds overhead and can be rather a bother. Then there's the hundreds of TB of data the model is dumping to disk. Now let's render that, shall we? Somebody call Pixar.

I am working on a project to simulation a thunderstorm which will produce a tornado in a "natural" way. The tornado needs to be adequately resolved. This simulations will have grid spacing of 10 meters. It requires a computer which hasn't been fully built yet (Blue Waters, in Urbana, google it). The time step will be 0.01 seconds, and the model will run for two hours of model time. It will take days of wallclock time. Keep in mind this model will have a physical domain not much bigger than about half the area of Oklahoma. Imagine global climate modeling now, and now you're talking 4 km resolution being all you can do.

This is why we need supercomputers to do high resolution weather simulations.

Old news

[jdb2]

The DOE as well as Oak Ridge, Los Alamos and Sandia National Laboratories already have programs in place to develop an "exascale" system by 2018. ( the date at which Moore's law predicts the possibility of such systems )
The top companies competing for the government funds are, not surprisingly, IBM and Cray.

See these two older /. stories here and here.

jdb2

Re:Computing for the next generation

[TheKidWho]

You've been simulated to die in our ongoing war with Eastasia, please report to the gassing chambers promptly to prevent the simulation from experiencing temporal improbabilities.

yeah, right.

[Major Downtime]

Pfff, old news. It will produce 42 as final output, and then we'll have it build another machine capable of performing one peta-quazillion calculations per second.

FLOPS, not FLOP

[91degrees]

You should realise that the "S" stands for seconds. Okay - it doesn't matter that much, but this is meant to be a technical site. The editors should really get this stuff right.

Bus innovation first, please

[BoldAndBusted]

What is really needed is faster *bus speeds*. So many CPUs just sit around waiting for data that sits across the bus. That's where the dramatic throughput improvements lie. Pretty please, DARPA? :)

obviousness for dummies

[epine]

The fact that the NSA is still serving a purpose in spite of 'completely secure' key sizes should suggest a fairly obvious conclusion.

Sweet. Stupidity by obscurity. Shall we integrate the area under the curve of obviousness * tinfoil_coefficient?

There is an obvious conclusion, but apparently it's not obvious. It's one of those cases where whichever answer you presently hold seems obvious, until one discovers an even more obvious answer. The parent post has been careful to distance itself from any clue as to which rung on the ladder of obviousness it presently occupies, a strategy which suggests an entry level rungs. Think of the cost. I certainly wouldn't want to be a large enough blip on the threat radar to find myself at the center of an exaflop computation. I value my keratin.

Feynman in Joking has a chapter on safe cracking. He ultimately concludes that "cold cracking" is largely a myth. Almost every safe cracker starts with an in: tampered mechanism, partially guessed combination, faulty mechanicals.

The bulk of what your average cyber TLA computes would be simple traffic analysis, which at that scale, is probably not so simple, and involves correlating across networks (cell, internet, house of poozle). One wonders how many initial demerits one earns by connecting through a known onion router.

Next you have attacks against keys with weak initial entropy, key leakage, or sloppy key management (betcha that's a growth industry). Any cipher which purports to send random bits can be hacked to leak key bits (secretly) in the apparently random nonce values. It's nearly impossible to prove your cipher isn't doing this without access to the source code all the way down to the CPU microcode, and beyond. Huh, a funny thing happened to our masks on the way to the foundry, but the chips seem to run great. From a TLA perspective, this is a useful advantage, because what you end up with is not a level playing field. What you can crack by brute force, someday soon your adversary can also crack by brute force. It's a lot more fun when you have to peel off the anonymous brown wrapper.

What seems obvious to me is that your average TLA enjoys hiding behind this obviousness meme, and might even participate in its dissemination as a part of a highly successful initiative in distracting paranoids and shallow thinkers from useful analysis. You just have to find a forum where seeming clever is more important than being clever, add water, and stir.

My favorite local coffee shop is right beside the schizophrenia resource center. If I had the right social hacking skills, I could accomplish this mission by buying the right person who drifts into the coffee shop with a wifi netbook a free coffee a day. "Just keep posting buddy, the Joe's on me."