3 ms·
I checked, the code they're running is ExaBayes. In my experience, centers like this are under pressure to find biologists to justify their systems, but most o
by dekhn 11y ago
I checked, the code they're running is ExaBayes.
In my experience, centers like this are under pressure to find biologists to justify their systems, but most of the bio user don't have codes that scale very much. THe "supercomputer" she's running on isn't very super.
Anyway, I used to be in the class of user, but I found that supercomputers never had the throughput to carry out real science. After arguing with DOE that they spent too much money on interconnect (and not really being able to run my codes because they "only" scaled to 1K cores) I moved to Google and built a system called Exacycle which runs embarassingly parallel codes (of which phylo tree generation is an instance; it's basically metropolis monte carlo sampling, although the MPI versions typically do some clever tricks to create long MC chains efficiently). Personally, I think our approach is far superior for all but the most tightly coupled codes- for example, our work doing huge collections MD simulations of proteins generated far more useful data to analyze than if you'd run just one long trajectory (this is a long-running argument in the MD field, but I think we pretty much nailed it in our paper, http://www.ncbi.nlm.nih.gov/pubmed/24345941 http://www.ncbi.nlm.nih.gov/pubmed/24345941
Frankly I don't see many computational science depts making more supercomputer-like services. It's hard to run these systems, they cost a bunch, and they are hard to keep utilized at a rate that is justifying their expense. Commercial clouds are eating away at them- as you say they are basically computers with shitty interconnect. But the reality is that their interconnect is "good enough" for nearly all bio codes I know of. It's far easier for a PI these days to pull out their credit card and pay for grad student cloud use out of pocket than spend time applying for and using supercomputers.
- stonogo 11y agoI'm glad Exacycle exists, but the tasks at which it excels are indeed not tasks for which supercomputers are usually employed. The things that are easy for a PI to pay for with Visa are generally not the things that are interesting to do with supercomputers. I see where you're coming from with the DOE stuff, though; those guys are all about high-visibility "status" computing, and I don't really find that scene appealing either. I'm not sure what you consider a supercomputer, if you don't consider a petaflops machine with a 3PB filesystem and >50Gb/s i/o "supercomputing." I guess we'll just have to disagree.
- dekhn 11y agoI used to work for NERSC when they ran Seaborg. Before that, i ran my PhD simulations on the T3E Mjolnir (PSC) and the other one, at SDSC whose name I forget. Back in those days, those were supercomputers, but a single cloud machine replaces one of those now. The only thing I consider a supercomputer is a device designed for capability computing where the capability is ability to run codes that exhibit strong scaling due to their interconnect. The petaflops you quote are calculated by the strong scaling of LINPACK due to its design. I think you need a large amount of petaflops these days to really be considered a supercomputer. And those petaflops come from having a combination of low latency and high bandwidth-- you couldn't just point at Exacycle and say "that's a supercomputer". The 3PB filesystem and 50Gb/s IO (you didn't say if that was per-node, or the bisection bandwidth of the central switch, or what) aren't particularly "super". In some sense this is just arguing about the definition of a supercomputer which isn't super interesting; I have a preconceived notion of what one is. If Amazon put a bunch of machines on an MPI switch (IIRC they actually have) and scored well in Top500, and you could buy time on it, and it scaled your code, they'd have one too.