Slashdot Mirror
Obama's New Executive Order Says the US Must Build an Exascale Supercomputer
Jason Koebler writes: President Obama has signed an executive order authorizing a new supercomputing research initiative with the goal of creating the fastest supercomputers ever devised. The National Strategic Computing Initiative, or NSCI, will attempt to build the first ever exascale computer, 30 times faster than today's fastest supercomputer. Motherboard reports: "The initiative will primarily be a partnership between the Department of Energy, Department of Defense, and National Science Foundation, which will be designing supercomputers primarily for use by NASA, the FBI, the National Institutes of Health, the Department of Homeland Security, and NOAA. Each of those agencies will be allowed to provide input during the early stages of the development of these new computers."
IEEE Spectrum had an article last year describing some of the challenges we'll need to overcome in order to achieve exascale computing.
Here's another, somewhat pessimistic piece they posted in 2008 - a digest of a DARPA report that went into significant technical detail.
The biggest hurdle is power, and the biggest driver of that isn't the actual computation (i.e., the energy to perform some number of FLOPS), but rather moving that data around (between cores, to/from RAM, across a PCB, and among servers). Other hurdles include how to manage so many cores, ensure they are working (nearly) concurrently, how to handle hardware failures (which will be frequent given the amount of hardware), and writing software that can even make use of such technology in anything approaching optimal fashion.
Not to say its impossible, merely hard given the present state of things and projecting a bit into the future. But as we know, "it is difficult to make predictions, especially about the future." [source?]
And random person freaks about because President exercises his lawful authority to tell agencies and departments under his jurisdiction to cooperate and present a plan for creating a supercomputer.
Here is a hint:
Sec. 7. General Provisions. (a) Nothing in this order shall be construed to impair or otherwise affect:
the authority granted by law to an executive department, agency, or the head thereof; or
the functions of the Director of OMB relating to budgetary, administrative, or legislative proposals.
(b) This order shall be implemented consistent with applicable law and subject to the availability of appropriations.
(c) This order is not intended to, and does not, create any right or benefit, substantive or procedural, enforceable at law or in equity by any party against the United States, its departments, agencies, or entities, its officers, employees, or agents, or any other person.
It is like nobody knows how the government operates any more, but if Obama does it, they're opposed, damn opposed.
No, but it can figure out you suck at the game.
Weather guys want this after NSA's done.
I'm a weather guy - running cloud model code on Blue Waters, the fastest petascale machine for research in the U.S. I don't think we've managed to get any weather code run much more than 1 PF sustained - if even that. So it's not like you can compile WRF and run it with 10 million MPI ranks and call it a day. Ensembles? Well that's another story.
Exascale machines are going to have to be a lot different than petascale machines (which aren't all that different topologically than terascale machines) in order to be useful to scientists and in order to no require their own nuclear power plant to run. And I don't think we know what that topology will look like yet. A thousand cores per node? That should be fun; sounds like a GPU. Regardless, legacy weather code will need to be rewritten or more likely new models will need to be written from scratch in order to do more intelligent multithreading as opposed to mostly-MPI which is what we have today.
When asked at the Blue Waters Symposium this May to prognosticate on the future coding paradigm for exascale machines, Steven Scott (Senior VP and CTO of Cray) said we'll probably still be using MPI + OpenMP. If that's the case we're gonna have to be a hell of a lot more creative with OpenMP.
A squid eating dough in a polyethylene bag is fast and bulbous, got me?
These Peta/Exascale supercomputers are build for computer simulations (climate change, nuclear weapons stewardship, computational drug design, etc.), not for breaking encryption. That's also one reason no one is using them to mine Bitcoins: they're just not efficient at that job. To compute lots of hashes, dedicated hardware designs (read: ASICS) far outpace "general purpose" supercomputers.
Computer simulation made easy -- LibGeoDecomp
What would the existence of an exascale supercomputer mean for today's popular encryption/hashing algorithms?
Nothing, nothing at all.
Suppose, for example that your exascale computer could do exa-AES-ops... 10^18 AES encryptions per second. It would take that computer 1.7E20 seconds to brute force half of the AES-128 key space. That's 5.4E12 years, to achieve a 50% chance of recovering a single key.
And if that weren't the case, you could always step up to 192 or 256-bit keys. In "Applied Cryptography", in the chapter on key length, Bruce Schneier analyzed thermodynamic limitations on brute force key search. He calculated the amount of energy required for a perfectly efficient computer to merely increment a counter through all of its values. That's not to actually do anything useful like perform an AES operation and a comparison to test a particular key, but merely to count through all possible keys. Such a computer, running at the ambient temperature of the universe, would consume 4.4E-6 ergs to set or clear a single bit. Consuming the entire output of our star for a year, and cycling through the states in an order chosen to minimize bit flips rather than just counting sequentially, would provide enough energy for this computer to count through 2^187. The entire output of the sun for 32 years gets us up to 2^192. To run a perfectly-efficient computer through 2^256 states, you'd need to capture all of the energy from approximately 137 billion supernovae[*]. To brute force a 256-bit key you'd need to not only change your counter to each value, you'd then need to perform an AES operation.
Raw computing power is not and never will be the way to break modern crypto systems[**]. To break them you need to either exploit unknown weaknesses in the algorithms (which means you have to be smarter than the world's academic cryptographers), or exploit defects in the implementation (e.g. side channel attacks) or find other ways to get the keys -- attack the key management. The last option is always the best, though implementation defects are also quite productive. Neither of them benefit significantly from having massive computational resources available.
[*] Schneier didn't take into account reversible computing in his calculation. A cleverly-constructed perfectly-efficient computer could make use of reversible circuits everywhere they can work, and a carefully-constructed algorithm could make use of as much reversibility as possible. With that, it might be feasible to lower the energy requirements significantly, maybe even several orders of magnitude (though that would be tough). We're still talking energy requirements involving the total energy output of many supernovae.
[**] Another possibility is to change the question entirely by creating computers that don't operate sequentially, but instead test all possible answers at once. Quantum computers. Their practical application to the complex messiness of block ciphers is questionable, though the mathematical simplicity of public key encryption is easy to implement on QCs. Assuming we ever manage to build them on the necessary scale. If we do, we can expect an intense new focus on protocols built around symmetric cryptography, I expect.
Note to ACs: I usually delete AC replies without reading them. If you want to talk to me, log in.