US DOE Sets Sights On 300 Petaflop Supercomputer

dcblogs writes U.S. officials Friday announced plans to spend $325 million on two new supercomputers, one of which may eventually be built to support speeds of up to 300 petaflops. The U.S. Department of Energy, the major funder of supercomputers used for scientific research, wants to have the two systems – each with a base speed of 150 petaflops – possibly running by 2017. Going beyond the base speed to reach 300 petaflops will take additional government approvals. If the world stands still, the U.S. may conceivably regain the lead in supercomputing speed from China with these new systems. How adequate this planned investment will look three years from now is a question. Lawmakers weren't reading from the same script as U.S. Energy Secretary Ernest Moniz when it came to assessing the U.S.'s place in the supercomputing world. Moniz said the awards "will ensure the United States retains global leadership in supercomputing." But Rep. Chuck Fleischmann (R-Tenn.) put U.S. leadership in the past tense. "Supercomputing is one of those things that we can step up and lead the world again," he said.

Comments

Nice and all, but where's the beef?

[DumbSwede]

I remember back in the 80's all the excitement about building faster and faster super computers to solve all sorts of grand challenge problems and how a teraflop would just about be nirvana for science. Around 2000 teraflops came and went and then petaflops became the new nirvana for science where we would be able to solve grand challenge problems. Now exaflop is the new nirvana that will solve grand challenge science problems once again. Seems raw computing power hasn't given us the progress in science we predicted. Sure it's been used for stuff, but it hasn't helped us crack nuclear fusion for instance, one of its often hyped goals.

Where's the score card on how much progress has been made because of super computing? I know drug design is one very useful application, but what are other areas that have been transformed?

Re:Nice and all, but where's the beef?

[chalker]

There are countless problems solved only as a result of supercomputers. Setting aside for a minute the minority of problems that are classified (e.g. nuclear stockpile stewardship, etc), supercomputers benefit both academia and industry alike. You'll be hard pressed to find a Fortune 500 company that doesn't have at least one if not multiple supercomputers in house.

For example, here is a list of case studies of specific manufacturing problems that have been solved http://www.compete.org/publica... which include things as mundane as shipping pallets, golf clubs, and washing machines.

The organization I work for, the Ohio Supercomputer Center, annually publishes a research report listing primarily academic projects that benefit from our supercomputers: https://www.osc.edu/sites/osc.... which range from Periodontal Disease, Photovoltaic Cells, Forest Management and Welding.

TL;DR: "HPC Matters" in many ways. Here's some short blinky flashy videos: http://www.youtube.com/channel...

Re:Nice and all, but where's the beef?

[JanneM]

How should one go about getting a job programming a large supercomputer?

Become a researcher in a field that makes use of lots of computing power, then specialize in the math modeling and simulation subfields. Surprisingly often it's quite easy to get time on a system if you apply as a post-doc or even a grad student. Becoming part of a research group that develops simulation tools for others to use can be an especially good way.

Or, get an advanced degree in numerical analysis or similar and get hired by a manufacturer or an organization that builds or runs supercomputers. On one hand that'd give you a much more permanent job, and you'd be mostly doing coding, not working on your research; on the other hand it's probably a lot harder to get.

But ultimately, why would you want this? They're not especially magical machines. Especially today, when they're usually Linux based, and the system developers do all they can to make it look and act like a regular Linux system.

If you want to experience what it's like, try this: Install a 4-5 year old version of Red Hat on a workstation. Install OpenMP and OpenMPI, and make sure all your code uses either or both. Install an oddball C/C++ compiler. Access your workstation only via SSH, not directly. And add a job queue system that will semi-randomly let your app run after anything from a few seconds to several hours.

Re:Ehhh Meh

[Macman408]

There are plenty of things that can use all the computing power you can throw at it these days. As you mentioned, weather forecasting - though more generally, climate science. Somebody from one of the National Labs mentioned at a college recruiting event that they use their supercomputer for (among other things) making sure that our aging nukes don't explode while just sitting in storage. There are thousands of applications, from particle physics to molecular dynamics to protein folding to drug discovery... Almost any branch of science you can find has some problem that a supercomputer can help solve.

Additionally, it's worth noting that these generally aren't monolithic systems; they can be split into different chunks. One project might need the whole machine to do its computations, but the next job to run after it might only need a quarter - and so four different projects can use the one supercomputer at once. It's not like the smaller computing problems end up wasting the huge size of the supercomputer. After all, many of these installations spend more in electricity bills over the 3- or 5-year lifetime of the computer than they do to install the computer in the first place, so they need to use it efficiently, 24/7.

Re:Ehhh Meh

[fahrbot-bot]

The number of floating point operations (FLOPS) performed by a next-generation game console outranks early days supercomputers like the Cray.

Sure, but do they have the system capability / bandwidth to actually do anything with those numbers and is their raw speed offset by not being vector processors like the Cray 2 (process an entire array of data in 1 instruction)? I'm not a hardware geek, but was an administrator for the Cray 2 at the NASA Langley Research Center back in the mid 1980s and, among other things, wrote a proof-of-concept program in C to perform Fast Fourier transforms on wind tunnel data in near real time - probably would have been faster had I been a FORTRAN geek - and the system could pump through quite a bit of data - at least for the 80s.

And the Cray 2 was way prettier than a PS3/4 or Xbox, though the Fluorinert immersion used for cooling is a bit cumbersome and expensive :-)