Slashdot Mirror
The UK's Fastest Supercomputer
bmsleight writes "The Guardian has a story on the HECToR, The largest supercomputer in the UK — around five times more powerful than its predecessor, HPCx, which is also at the University of Edinburgh. It measures up well internationally, sitting at 17 in the top500.org list of the most powerful computers in the world."
I have really been impressed with the level of commitment to science, research and education outside of the US right now and efforts like HECToR only consolidate that impression. While we here in the US have essentially dropped the ball on education and science funding for the past oh, six or seven years, the rest of the world is really stepping up. Of course I have mixed feelings about this as I am a US citizen who works in science and education, but it is also good to see other countries stepping up. For instance, a few months ago, I visited the University of Leicester and was truly impressed with the focus and quality of the research going on in the UK. Their commitment to bioscience funding is something that the US government should be very careful about as we stand to lose some valuable talent overseas if we are not careful...
Visit Jonesblog and say hello.
The UK GDP is 5th in the world (nominal) or 6th in the world (purchasing power parity). If our best supercomputer is coming in at 17th, we aren't spending enough on research.
Not to belittle this project, of course, building the worlds 17th fastest supercomputer is an achievement in anyone's book - but it is a sign of where the UK government is weak.
If we can put a man on the moon, why can't we shoot people for Apollo-related non-sequiturs?
It measures up well internationally, sitting at 17
The British don't mind being at any number as long as the best French *whatever* is lower ranked - 19 in the case of the latest supercomputer list. Although they might be a little out of sorts that Spain is above them at 13.
Note: if you are British or have any British friends, the above is 'funny' or 'insightful', not 'flamebait' or 'troll'.
...the answer to Life, The Universe, and Everything?
GetOuttaMySpace - The Anti-Social Network
The article really didn't say much about HECToR itself. It's a 60 cabinet Cray XT4 system that currently has over 5500 AMD dual core processors. We'll be upgrading it in stages over the next couple of years to over 250 Teraflops. Including some cabinets of the new Black Widow Vector product, now called the Cray X2 system. The Cray team, myself part of it, is actually a multinational effort. I'm a US citizen who is headed over to maintain the system, we have a Brit on the team and the third is also from outside the UK. It's an interesting situation. The biggest UK system, being maintained by two expats and a local. (-: ice_hawk55
The people who truly understood how our Nuclear stockpile worked are all older and retiring. For years and years it's not been a field of study that was popular with Phd students. Now we're in the situation where the people who know what's going on are retiring and there are not enough new folks coming down the line that understand what's going on. Add to that the fact that you can no longer actually set one of these beasts off and they have been sitting idle for decades. What's the state of the current stockpile? The only way to predict how these decades old weapons will react is to simulate them with a supercomputer as best you can.
Doesn't this Top500 contest boil down to a matter of who has more money than the other?
I mean, at this stage, there isn't any real innovation in interconnect or processor or memory technology. It is mostly a matter of who has the money to buy thousands of these chips, cobble them together and supply enough money to keep the whole thing running.
If University of Edinburgh had thrice the money, they could cobble three Hectors together and then they would have had a system at least twice as powerful or may be only 50% more powerful (Whatvever the power gain is). Then they would end up higher on the list.
May be there should be some kind of constraints built in within the Top500 to encourage actual innovation as opposed to measuring the financial resources of an institution or a country.
Your example of the 2400 baud modems for a linux cluster isn't completely accurate, as linpack does a little bit of communication, though the point is well taken. The top500 list only uses linpack to measure performance, and linpack represents a very easy problem to solve. Essentially, the top500 list is a list of which machines do a really good job of solving a trivially difficult problem. The hpcc benchmarks (http://icl.cs.utk.edu/hpcc/) are a lot more interesting; though, even these need to be read with some caution.
Ranking supercomputers is a really hard problem. Each application has different needs for communication latency, bandwidth, programming model, cache size, memory bandwidth, and computational throughput. Then you have to ask: how much optimization can I do to the benchmark? Am I going to be able to do the same amount of optimization for each of my applications? How easy is it to extract this performance? The guys writing the software for these things are usually professors or post-docs in the hard sciences, not in supercomputing.
If you want to see the pictures, here they are.
160 GB Hard Drive vs RAM. Not a good comparison. They didn't mention the SAN storage we have, the tape backups, etc etc. I know. I know. I'm being picky.
Except it's not doing research. It's eavesdropping on all electronic communication passing our borders. Welcome to 1984, say hi to Big Brother.
Money for nothing, pix for free
The CPUs are arranged in a Torus shape, according to here. I've seen a lot of these parallel computers with this shape. I can't think of how to make Google tell me this, so perhaps someone here could. What is it about the torus that makes it a good shape for this situation? Have other shapes been tried?
I have the feeling that an arrangement where the connectivity of vertices (CPUs) was distributed according to a power law (i.e. a few vertices with lots of edges, most with not many at all) would minimize the distance between any vertex. I don't think a torus gives you that. Maybe I'm looking at it the wrong way though.
How dare you be so modest!! You conceited bastard!!