The 'cheap' supercomputer in the Amazon cloud
$4,829 per hour isn't "cheap" but a 50,000 core supercomputer would cost up to $25 million.
A cancer research software simulation project run by New York firm Schrodinger relied on 6,742 Amazon EC2 instances to provide 51,132 computer cores. This global effort used Amazon installations in Virginia, Oregon, California, Tokyo, Singapore, Sao Paolo, and Ireland. A total of 58.78TB of RAM was used in the "computer" for the project.
By winning a contest run by Cycle Computing for free supercomputer time, Morgridge Institute for Research was able to perform stem cell research. Existing supercomputer owners do sell time on their systems, but getting 50,000 computers working together takes careful planning, and will not be done regularly. Until, of course, Amazon EC2 scales up some more.
Buy versus rent
It's much cheaper to just budget $100k and book a few weekends with this thing.
jeblucas on arstechnica.com
It ran for only 3 hours, at a total cost of $14,486 - compared to the build cost and lead time involved in building a "$20-25m data center". This has huge implications for the availability of supercomputers for smaller organizations and use cases.
garethsprice on news.ycombinator.com
Our department which currently runs a 1000+ core machine and 2500+ core machine for MD simulations is still waiting on jumping onto the Amazon/Cloud bandwagon mostly because it's still not cheaper than owning a cluster for a few years.
michaelgrosner on news.ycombinator.com
If the calculations were so parallel, and it took 50k cores 3hrs, wouldn't it take their home cluster of 1.5k cores ~100hrs (a holiday weekend)?
FirsbeeFreek on arstechnica.com
Long row to hoe
Having been privy to the state-of-the-art projects in a lab with a ~5,000 node cluster it was still disappointing to see how rough the whole cell modeling approaches were.
jboggan on news.ycombinator.com
Individual wet-bench work is good to use as prior data, but it's just a hint at a part of a large, complex system, and overly reductive approaches are going to completely miss the big picture.
epistasis on news.ycombinator.com
A single CPU used for the project would take almost 13 years to run the same test.