In this podcast, Rich Brueckner from insideHPC interviews Steve Scott, CTO of Cray Inc. Scott recently returned to the company after a few years at Nvidia and Google.
insideHPC: You’ve just come back to the company and you’re CTO once again. What drew you back to Cray?
Steve Scott: I went off and spent a year at Google. Google is a great company, but I really said my heart was in HPC. I love working with the customers. I love the direct applicability of what we do to really important societal problems. I just realized that’s where my roots are and that’s where my passion lies.
insideHPC: Congratulations, by the way. Now you’ve seen clouds from both sides. Do you have a new view of the world now that you’ve been away?
Steve Scott: Google is a very different environment. They’re obviously getting big into cloud computing as are Amazon, Microsoft, et cetera. Cloud computing has its place. I think we’ll start to see some high performance computing workloads moving to cloud. But cloud computing, I don’t think, can or ever really will replace a dedicated HPC infrastructure. The machines just aren’t optimized as much. The cloud computers are not going to be as efficient as having a dedicated HPC system.
insideHPC: So you were working on these systems years ago, you’ve come back. Cascade is out now and doing well. Is that gratifying to see that all coming together?
Steve Scott: Yes, it’s pretty great. I felt like a proud papa watching Cascade from the outside world and it’s doing really well. It’s pretty incredible. I have to say I’m really excited by what I’ve seen since coming back to Cray. I’ve only been back a week, and I spent last week at our principal engineer workshop offsite hearing all about the things going on in the company. I’m really jazzed about all the things that are going on at Cray.
insideHPC: I know you’ve only been back a week, but I’m kind of curious. What’s first order of business for you? What’s the big tough nut that they’ve asked you to come figure out?
Steve Scott: My first order of business is just getting up to speed with everything that we’re doing. I’m really excited about the fusion of traditional high performance computing and data analytics. As I look at what’s going on now, we’ve got this tremendous influx of data, no surprise everyone’s talking about it. All that data needs to be processed, and it’s getting up to the scale now where you really start needing high-end HPC systems that have traditionally been designed for processing scientific data and simulations, to process all that data. There is a virtuous cycle there, because then as you apply more complex computational algorithms to the data, you can do more complex models, that allows you to look at more types of different data streams, and it goes round and round.
We’re also starting to get some real traction in data analytics, both on the traditional data analytics side, and also on graph algorithms, and some of the things that we’re doing with our YarcData and Urika products. Seeing that the capability of taking our standard HPC systems and the data analytics, merge them together, and work on problems at the boundaries of those, the sort of disciplines, I think is tremendously exciting.
insideHPC: I can hear the enthusiasm in your voice. This is terrific to have you back. Welcome back.
Steve Scott: Thanks very much. Feels like coming home.