Supercomputing as a Service, Cheap


by Tim Moran

This list of the Top 500 Supercomputers comes out every June and November, and I doubt if the minor ranking variations from one list to the next hold much interest to those not in the supercomputer game.

Most enterprise IT folks have little to do with these processing behemoths. And there's not a lot of drama to the rankings, either. In fact, according to TOP500.org, the keeper of the twice-yearly tabulation, the same 10 machines topped this edition as the one from June.

But further down on the list at #42 sits Amazon Web Service's self-made EC2 Cluster Compute Instances (Xeon 8C 2.60GHz, 10G Ethernet), built on a 17,024-core, 240-teraflop cluster. What makes this so interesting is that Amazon is bringing high-performance computing (HPC) to the masses--at a very elite level.

Says Amazon: "HPC allows [users] to solve complex science, engineering and business problems using applications that require high bandwidth, low latency networking, and very high compute capabilities. . . . Using Amazon EC2 Cluster instances, customers can expedite their HPC workloads on elastic resources as needed and save money by choosing from low-cost pricing models that match utilization needs." And how does being able to do all this for $2.40 an hour sound?

There is certainly nothing all that new about being able to tap into computing power such as this--it's been done in science and engineering for quite some time (although not without difficulties). But Amazon has created an on-demand HPC environment at a cost that begins to make it viable to the enterprise.

As Big Data moves more and more into the mainstream of business--ask your marketing execs about it if you've any doubts--having computing options like this, at a reasonable price, could very well change the way IT thinks, plans, and works.

Or as Jeff Barr, author of the Amazon Web Services blog wrote: "What can you do with a supercomputer of your very own?"