CNET News Video
High-performance computing for Web 2.0At the Web 2.0 Expo in San Francisco, Sun Microsystems CEO Jonathan Schwartz talks about the company's new high-performance computing facility in Austin, Texas, and how Web 2.0 companies like Facebook and Google are benefiting from high-performance computing...
^B00:00:00 >> What a really interesting insight, seeing your presentation at this Sun Analyst meeting was -- when you talked about the HPC market as not actually just being science. >> Right. >> You know, that really is this world we're talking about where HPC is bleeding -- HPC being high performance computer -- bleeding into everyday life and business applications, and in fact, all these new next-generation applications. >> Well, so the high performance computing facility Tim is referencing, we just opened in Texas. It's the Ranger Facility at the Texas Advanced Computing Center. It's a 500-teraflop system, so 4,000 blades, 16,000 cores, terabytes of ram, you know, pedabytes of data, 200 terabit switches with Texas longhorns on them in the middle of the facility. And what was interesting, when we were opening it, is in the audience --you know, this was opened by the NSF. This particular HPC facility is bigger than all other NSF granted high-performance computing facilities combined. So it's really quite impressive. And it's an open facility. People can submit jobs and, you know, and there were a couple things that were interesting about it. One is within two quarters of its existence, they were out of capacity because the requests for usage of the system had exceeded what they had available. So they had to start rationing again, which was interesting. But the other thing that was interesting is in the audience were IT decision makers for drug companies, energy companies, financial services companies... >> Yeah. >> ...because they're looking to high-performance computing for business advantage. >> So, you know, clearly there's high-performance computing. There's a world of traditional high -- high performance computing applications. And then there are applications like Google and Facebook, who tend to be running on huge clusters of commodity PCs. >> You know, you might think that until you go hang out in their data centers and you realize their average node is now a four way quad core. And to me that looks like a 32 way computer, and... >> Uh-huh. >> You know, so I think -- and by the way, when you sit down and talk to folks at companies like Facebook, they start talking to you about high-performance computing to interpolate and interrogate the social graph. And they all of a sudden need terabit switching and -- so I think we're seeing a very, very interesting shift toward how do we simply serve the web to how do we run analytics against it? >> Right. So you think the markets going to come back in Sun's direction from the [inaudible] commodity market? >> I -- there's no doubt in my mind. I mean I see it every day. And there comes a point, especially -- virtualization's a good example of it. >> Yeah. >> Where all of a sudden you say tell me why I have 32 independent machines when it would just be easier if I had it all in one place. >> Yeah. >> You know, the only problem, historically, with SNP machines is they were frightfully expensive. >> Yeah. >> You know, if you make it less than 32 times the cost of a one-way computer, all of a sudden it becomes less expensive. ^E00:02:51