Getting compute time on a supercomputer has long been a privilege limited to physicists, scientists, medical researchers and other academic types.
Researchers had to apply for time by submitting project proposals in hopes they would be deemed worthy of approval and priority access to a supercomputer. In fact, getting time on a supercomputer used to be as competitive as getting time on the world’s most powerful telescopes. There were long waiting lists. But that’s not necessarily the case any longer.
According to reports, Amazon has built a virtual supercomputer that runs on Amazon Web Services Elastic Compute Cloud (EC2) and is looking to change the supercomputer access paradigm. Amazon has named its virtual supercomputer the Elastic Cloud Computer and it is ranked as the 42nd fastest supercomputer in the world. Amazon’s Elastic Cloud Computer delivers some 240 trillion calculations per second, or 240 teraflops on 17,000 cores.
While that is a rather impressive feat, it does fall far short of the current supercomputing champion, Fujitsu’s K Computer, which maxed out at 10 petaflops (10 quadrillion calculations a second) in November 2011. Nevertheless, the Elastic Cloud Computer proves to be a comparative bargain, almost anyone can use the system for $1,279 per hour, or $11 million a year if run full time.
Fujitsu’s K Computer costs Fujitsu around $10 million per year just for the power bill. Fujitsu’s published specs also states that the K Computer cost around $20 million to build and consists of 864 racks with 88,128 interconnected CPUs. Each processor in the K Computer is linked to 16 GBytes of RAM, bringing the memory total 1,377 terabytes, which requires 9.89 megawatts of power, about the same as 10,000 suburban homes.
By democratizing access to supercomputing, Amazon may very well change the business intelligence and big data markets for many businesses, which will now be able to run advanced algorithms on large amounts of data to identify trends and build plans of action based upon big data analytics.
Prior to the Elastic Cloud Computer, businesses either had to build their own super compute clusters at a cost of tens of millions, or rent physical access on an existing supercomputer, if available and a price dozens of times Amazon’s hourly rate.
Home Applications