Late Friday afternoon I had my first phone meeting with Mark Barrenechea, CEO of the new SGI (it’s pronounced “baron shay”, by the way). You will no doubt recall that SGI was created by the acquisition of Silicon Graphics (nee SGI) by Rackable Systems. Barrenechea is one of the blogging CEOs (read his blog at ceoblog.rackable.com); and although not as prolific as Sun’s Jonathan Schwartz, he may well soon be HPC’s only blogging CEO. He was affable, articulate, and confident — some of SGI’s employees have described him to me as an engineer running an engineering company — and knowledgeable about our community.
Barrenechea may be new to high end HPC, but he isn’t new to tech. He became president and CEO of Rackable Systems in 2007 after joining that company’s board in 2006, and prior to that he was Executive Vice President and CTO for Computer Associates, reporting directly to CEO John Swainson. Before that he was Senior Vice President of Applications Development for Oracle Corporation, and a direct report to CEO Larry Ellison. Altogether, he’s spent 20 years in the tech industry, and interestingly has two books to his credit: eBusiness or Out of Business and Software Rules.
Barrenechea started the conversation by offering, before I could get to the question myself, to explain “why I did what I did.” Rackable has been very successful building scale out clusters at a very large scale. Barrenechea explains that these aren’t cookie cutter systems, but built to order for his various high-end enterprise datacenter customers. These solutions leverage Rackable’s “secret sauce”: cabinet-level enclosure technologies. All of this together, according to Barrenechea, “has allowed [Rackable] to grow up to be larger than Sun in North America in x86 compute, while providing our products into some of the largest internet properties, whether they be serving up a billion videos a week, managing some of the largest search sites, or social networking platforms.”
This isn’t hype. If you go check out Rackable’s customer page you’ll find internet names you and your kids know well: Yahoo!, Amazon, YouTube, CraigsList, Facebook, and others. Rackable and rival Verari had this business largely to themselves for years, but recently IBM, HP, Sun, and Dell have all decided they wanted a piece of the scale out pie. This competition, and a horrific financial environment, have led to falling financial performance for Rackable in recent quarters and provided additional incentive for Rackable’s executive team to look for new ways to round out their business.
The company had an HPC business of sorts, but it was aimed at “do it yourself” customers with their own expertise and software stack who were only looking for hardware to install it on. “But,” says Barrenechea, “in the broader market, where we needed to be experts in scale up and scale out, it was taking us too long on our own to do it.” The vision is that putting the two companies together preserves the business Rackable has already built in scale out while providing additional markets in federal and commercial high end computing with a technology channel that runs between the two businesses. “I want to create a company that is seen as the expert in large scale out solutions, HPC, and visualization,” says Barrenechea.
Here’s a fundamental question, though: how do you put two struggling companies together and come out with a successful company? Barrenechea points to SGI’s longstanding R&D investments on the plus side, and to problems in that company on the business side (such as the debt payments that resulted from the previous bankruptcy filing), a deep recession, and operating inefficiencies like too many facilities and supply chain problems. In terms of Rackable, Barrenechea points out that while revenue has been down, the company has managed to increase cash on hand; it is this cash that was used to fund merger. He expects that Rackable will be able to bring better supply chain management to the Silicon Graphics part of the business, and that combining the two companies will allow them to share operating costs and get better economies of scale on the operating side of the new company.
I don’t run a public company for a living, so my opinions here don’t matter all that much. I do know, however, that these are the same kinds of benefits that every executive announces following a merger. It doesn’t take much Googling to figure out that although the guys at Ivy League school A might be quibbling with the guys at west coast powerhouse school B over the precise rate of failure of mergers and acquisitions, the vast majority of them do fail. Some really big examples? Daimler and Chrysler, HP and Compaq (thousands laid off, Dell gained market share, took four years to even out), and AOL and Time Warner (B school case study on how not to do mergers, now rumored to be devolving after destroying much shareholder value). Does this mean the SGI acquisition is doomed? No. But it does mean that the expected benefits are likely to not materialize as expected, and that Barrenechea and his team are going to have to be creative and flexible if they are going to make it work.
Before the financial performance is measured, though, there has to be the actual, day-to-day merging of two different companies. Can a scale out, Web 2.0 focused company really merge with a high-end HPC company and make it work?
Barrenechea says yes, at least based on his own experiences. “I look at my 8 years at Oracle working directly for Larry Ellison. You had multiple cultures within that company. You had the applications approach to the world. You had the database approach to the world. And you had a set of tools, and then professional services that wrapped around it. What made it work was that they were all related: they weren’t unrelated technologies.” He has a point: both markets are worried about issues like density, power distribution, and qualifying the latest chips. And he seems to recognize where the differences are (for example, in HPC applications expertise) and to know which part of his new company has the right experience to deal with those differences.
What about SGI’s technologies and its road map? One of the most significant items on the road map for the old Silicon Graphics was Ultraviolet (UV), the planned hardware platform that brings hardware-enabled shared memory onto Intel’s mass market Xeon platform. Will UV survive? “If you don’t believe in UV, you would not have brought the two companies together,” says Barrenechea. “We are fully committed to UV, and it is paramount to our future.”
At the same time Barrenechea was at pains to explain that SGI is fully committed to the current generation of shared memory products based on Itanium, and said that although there isn’t a final decision the company’s current thinking is that SGI will continue to offer shared memory systems based on both chips. I’ll be interested to see what market there is for Itanium shared memory after Xeon-based systems start showing up. My own opinion is that there are so few applications in which Itanium will make enough of a difference that customers are willing to shell out the extra Benjamins that this product strategy will not survive.
Under CEO Bo Ewald’s leadership the company got back into visualization, and had just recently announced its VUE line of products (PowerVUE, RemoteVue, and FusionVue). Does visualization survive at the new company? Barrenechea says yes, but he describes this as a technology that is in “incubation.” I’ve spent some time with the VUE line, and I think it has potential, but Silicon Graphics didn’t seem to have a clear picture of the market for the product. Does Barrenechea have a better idea of where VUE fits into a viable market? Yes, but he’s not telling us. “I have a strong point of view, but I’m not ready to announce it except to say that it’s important, we’re committed to visualization…and watch this space.”
My colleague Mike Bernhardt, who’s been in HPC marketing for 20-plus years and so knows something about the stickiness of companies and products, has a take on the renewed SGI that feels right to me:
We’ve seen our share of companies with great promises — and even great technology — that just couldn’t find the staying power. To date, the way I see the history of this community, the end-user decision makers have been very forgiving of SGI, have been willing to work with them under the most bleak conditions, and anyone in this community who’s been knocking around for a few years has to admit that SGI has staying power. Let’s revisit this again in November, when it is once again SGI’s time to stand under the spotlight at SC09. I for one would certainly like to reserve judgement until then.