Jason Stowe, Cycle Computing, at AWS Re:Invent 2013 2013 with John Furrier and Dave Vellante Jason Stowe, CEO, Cycle Computing, explained the company's latest AWS use case which leveraged "Amazon Spot instances to create the most cost effective supercomputer in the world." Stowe discussed this impressive results with theCUBE co-hosts John Furrier and Dave Vellante, live at the AWS re:Invent conference. The use case in question, explains Stowe, ran 2.3 million compute hours in 18 hours. Cycle Computing used about 150,000 cores to power research focused on finding materials to turn sunlight into electricity more efficiently. The cloud-based supercomputer "would have been worth $68 million to buy it." The total infrastructure cost with Amazon was closer to $30,000. Asked how long it would have taken to provision the hardware for this use case, Stowe said it would have taken 12 to 18 months. The supercomputer was used to help Dr. Mark Thompson, professor of chemistry at the University of South California, to test more efficient materials to turn sunlight into electricity, finding materials that are cheaper to manufacture and highly efficient. "We were able to take 205 of those materials and in 18 hours determine if they were efficient." As far as other supercomputer projects go, Stowe mentioned a 10,000 instance that would be soon presented to a customers, as well as the previous 50,000 core test with Schrodinger. As far as the physical infrastructure is concerned, he said "we help customers that have tens of thousand of cores in their infrastructure." Looking to the clouds for innovation Discussing cloud adoption in the HPC business, Stowe said that his company's clients -- which include top pharma, financial, and insurance companies -- are "all looking at cloud to get them innovating again. You want your smartest people asking the right questions," those that will change your business, or cure cancer. Cycle Computing is trying to create that infrastructure to allow them to ask the right questions, "not trying to fit in what we could afford last year." Asked what speed bumps the use case had highlighted and what he'd learned from it, Stowe said "the hard part for us was essentially resiliency and dealing with scale. Some of our test runs were larger than Fortune 500 companies." To manage that, "we wrote a scheduler, a way of distributing that job workload" called Jupiter. "We can get hundreds of thousands of cores doing millions of things," and if they shot down any machine or data center, it would still complete its tasks. AWS's utility model churns up the supercomputer Cycle Computing is the "leading software company in utility HPC," Stowe said, catering to government, financial, and health. "We write software that connects smart brains to supercomputer clusters very easily." The utility model that AWS provides is able to turn into a supercomputer at the push of a button. Asked to share his view on the new modern era of computing, Stowe said, "it's really simple. Every brain, every device, connected to whatever compute power is needed, to solve whatever problem needs to be solved." Commenting on research funding models, Stowe said, "that's actually where I think the Spot part deserves attention." He likened the cloud to insurance for your home. Infrastructure in the cloud meant "amortizing the infrastructure risk" over all cloud users. Spot makes compute power available to scientists and engineers and allows them to do work at a low cost. @thecube #AWSreinvent