Contrary to Misleading Rumors – Cray Continues Its Focus as a Systems Company

Print Friendly, PDF & Email
Intel and Cray should be applauded for smart, strategic business decisionA number of publications have voiced their opinions that Cray is shifting its focus and strategic direction to software. Some public comments have even gone as far as stating that Cray is becoming a software company.

The Exascale Report interviewed Cray CEO, Pete Ungaro, to set the record straight.

The bottom line: Contrary to many misleading articles, Cray is not rebuilding itself as a software company.

This interview also available as an audio podcast. http://dl.dropbox.com/u/8413402/Ungaro%20Interview%2005-11-12-final.mp3

< < < < <> > > > >

The Exascale Report: Does the recent sale of Cray’s interconnect hardware assets mark a strategic shift in Cray’s business model and technology direction?

Ungaro: That’s a great question Mike. In no way does this change our strategy or direction. We’re still very much focused on being the best supercomputing company on the planet, with very differentiated technology, and then work to leverage that technology into new markets – like storage and big data. We’ve always been a systems company, and from my perspective, we always will be a systems company.

TER: So, as far as a change to your roadmap for future technology and system announcements – storage and big data – can you expand on that?

Ungaro: You know, one of the nice things about this agreement is that we continue to have total control over our roadmap. So, our current products, XE6, XK6, those supercomputers are completely unchanged. Our next generation supercomputer and the strategy around that, which we call Cascade, is unchanged, and of course our storage and big data products, Sonexion and uRiKA, are also unchanged from this agreement. The roadmap over the next handful of years is really all unchanged. And then off into the future of course, we’ll leverage other interconnect hardware from partners like Intel, but maintain our differentiation, and our system level differentiation, to get leadership performance at scale, which is really what people come to us for.

TER: So, could you expand on that just a little bit, because I have to admit, when I sit back and look at this, I don’t understand really how you will differentiate.

Ungaro: Yeah, that’s a fair question. Before this ever happened, we always talked about our differentiation in three kind of big buckets. One is all around the interconnect hardware and software, so the ASIC that we developed that was the interconnect ASIC as well as all the software around optimizing that at the system level. The second area that we focused on was software to manage these huge supercomputers and get as much performance out of them as we possibly could. And then the third area is around power packaging and cooling technology – really those technologies that we do to densely package our machines to drive down that total cost of ownership by making them very energy efficient and upgradeable over time. So, as we go forward now with this new agreement, out of all of those areas, the only area that we are giving up is part of area number one, so for the interconnect hardware and software side, we’re not going to do the hardware any longer. We’re continuing to do the software, but we believe that with the trends in the industry, that doing the hardware out into the timeframe that we’re talking about, which is out past our Cascade system, just isn’t the right business or technical decision for the company.

And so, we’ll still do all the systems software to pull all the optimizations out of interconnect , and part of our partnership with Intel was a collaboration to allow us to do that. And then of course all those other areas, software for system management for performance, technology around power packaging and cooling, which are hardware technologies by the way, all of those remain the same.

TER: So, listening to what you’re saying, it seems like a bit of an exaggeration for someone to define Cray as becoming a software company.

Ungaro: Yeah, it’s definitely a bit of an exaggeration. Like I said, before and after this, I’ve always viewed us as a systems company, and being one of those companies willing to invest the time and energy to make the integration of software and hardware come together and deliver performance at scale.

You know, we’re not going to be one of those companies that are kind of a ‘rack and stack’ cluster builder. There are plenty of those companies out there in the market today and I personally don’t think the HPC community needs another person that does that. What Cray’s claim to fame has always been, and in my mind, will always continue to be, is a company that is very differentiated and works at bringing things together at the system level , especially at scale, and especially at the highest sustained performance possible.

TER: Can you give us a glimpse of how your revenue base might change over the next 3, 4 or 5 years if you think of a balance of hardware and software and services?

Ungaro: You know, it’s interesting. Even today we don’t break our revenue out that way. The way we have typically broken our revenue out is products and services, and we bring hardware and software together, so we don’t really, more or less, charge for a lot of our software. The technology is all integrated at the system level – and that’s just a part of how we think about things as a company which is at the system level. Even if you look at our forays into things like big data with uReKA, you see a very integrated – almost appliance-like architecture that we are building for that environment.

Today, we’re about 20 to 25% services and the rest is product. If we look at things playing out over the next few years, we plan on adding to that with our big data effort, adding to that with our efforts in storage, and adding to that with our efforts in what we call mid-range supercomputing , but I don’t fundamentally think that those ratios are going to change all that much.

TER: So, looking at your product development after you complete the current range of products – let’s say 3-4 years out, will you guys be getting locked in to Intel processors at that point?

Ungaro: Definitely not. One of the things that is part of our agreement is that we maintain control of our roadmap, and that means we still plan to partner with our long term partner vendors like AMD and NVIDIA and other partners, so this agreement in no way ties us to Intel. Although, I think as you and many other folks out there are aware, our next generation system, called Cascade, also allows us to integrate Intel processors into it. So, we’ve had this vision where we build a single system and we are able to put different processors and technology into that – we call that adaptive supercomputing – and that’s still our vision going forward.

TER: So Pete, do you still also believe that the path to exascale requires heterogeneous systems?

Ungaro: Yeah, absolutely. I think that’s clear, at least in my perspective and many other people out there in the industry. I think the question Mike is where will the heterogeneity be at? Is it going to be at the silicon level? The socket level? The node level? Or, at the system level? I think there’s still a lot of research work to figure that out, but I definitely believe the kinds of systems that we’re going to be building, especially dealing with both simulation and analytic environments, the combination of supercomputing and big data, I fundamentally believe are going to need heterogeneous systems.

TER: Having personally been involved in a number of mergers and acquisitions and technology buy-outs in my past, I’m very curious from the human relations standpoint, how this thing mapped out. What can you share with us Pete in terms of how long have you been working on this arrangement and how did this play out for the employees who were involved?

Ungaro: We’ve been of course working with Intel as a partner for many years – since we’ve been planning our Cascade system and our future systems, and thinking about exascale and such, and while we’re not disclosing dates and other details, I would say that Intel came to us a little while ago with their plans and what they wanted to do, and from that this was kind of the outburst from all of that.

You know, as it worked out – from a technical perspective, it was the right technical decision for us – from the marketplace, and from a business perspective, it was the right business decision, so it was a really easy decision from that side, but as you correctly say, the more difficult thing was the employee side. I would say it was the most difficult part of the decision – letting 74 people in the company who have been amazing employees and have done amazing things for Cray – move on to a new company and on to a new career. So, that was clearly the trickiest part of it.

TER: And has that transition already taken place?

Ungaro: Yes. We notified our employees right before it was publicly announced, so they knew jus t slightly ahead of when everyone knew, which as a public company, with a material transaction such as this, we’re very careful about that. 100% originally accepted to move to Intel . When we closed the transaction on May 2nd, those affected became Intel employees. We actually worked out with Intel arrangements of where they were going to sit and we opened up parts of our facility. We wanted to provide a really smooth transition for all the employees moving over to Intel. And Intel has done an amazing job of helping them through that transition.

TER: How do you feel about the community’s reaction to this announcement and what you have seen so far?

Ungaro: It’s been interesting. I think a little bit at first – people were shocked. I’ve been able to spend a lot of time with our customers. We had our annual user group meeting last week in Germany, and as you can imagine, I talked to a lot of people on the phone, as did many others at Cray. People seem to really understand it. When they think about where the technology trends are going toward the end of the decade, and once they realize we have total control over our roadmap – with things like our current products and even our next generation product, Cascade, that we won’t change those roadmaps, we’ve gotten really good, positive feedback – people being really excited about the opportunities this opens up – both for the companies as well as a stronger and closer relationship between Cray and Intel. It’s been really positive. I’ve been really pleased about it. I would say the same on the employee side. People really got it and really understood it. It was a much smoother transition for both employees and customers than I had originally imagined – and I’m really happy about that.

TER: Well, it sounds like a very smart business decision for both companies so I congratulate you on that.

Ungaro: Thanks Mike.

TER: Well, that’s all I had unless you have something else?

Ungaro: No, I think you covered a good perspective of it. I would just say – maybe the thing to leave with is – I think it’s something where the company is really creating the future. When you’re in a high-end technology business, you have to look at where technology is going and you have to move ahead of the market. And I think we’ve done that really well as a systems company. Back a few years ago, we were even building our own vector processors – and we made the decision to move off that and I think that was one of the smartest decisions as a company that we’ve ever made, and we were able to invest in other technologies that have longer and more sustainable differentiation in the market, and we’re going to repeat that here yet again and I feel really great about it. I think you and the rest of the market should expect some pretty exciting things out of Cray in the years to come.

For related stories, visit The Exascale Report Archives.

We’ve always been a systems company, and from my perspective, we always will be a systems company.

 

So, we’ve had this vision where we build a single system and we are able to put different processors and technology into that – we call that adaptive supercomputing – and that’s still our vision going forward.

Comments

  1. JuneGuest says