Features:

HP VETERAN STEVEN JOACHIMS JOINS OCTIGABAY: INTERVIEW
by Alan Beck, Editor-in-Chief

On September 22, 2003, OctigaBay Systems-a new entrant to the high performance computing arena-announced the appointment of Steven Joachims to the role of vice president of business development. Joachims joins OctigaBay with 20 years experience in high performance computing, most recently from Hewlett-Packard, where he was the director of marketing and solutions for HP's worldwide server business unit in high performance and technical computing markets. HPCwire conducted an interview with Joachims to get his perspective on changes in the high performance market, both past and future.

Interview

HPCwire: Why did you leave HP for a start-up like OctigaBay?

JOACHIMS: I had a wonderful experience at HP and it was difficult to part with such a great group of people. However, when I was offered the opportunity to join OctigaBay, I knew it was something I couldn't pass up. Not only is OctigaBay on the cusp of delivering something truly innovative, we have a leadership team with an impressive background in bringing new innovations to market. I believe the HPC market is ripe for innovation right now, and I'm to help lead the company in executing on that vision.

HPCwire: Can a start-up be successful in the high performance computing space, an area dominated by large commercial vendors?

JOACHIMS: Absolutely. This is the third start-up I've been with in my 20 years in high performance computing. I had the opportunity to join Convex Computer Corporation at a very early stage and to be part of creating a new market space that became known as "affordable supercomputing". Convex was ultimately acquired by Hewlett Packard, and we still see much of the core Convex technology in the HP Superdome family today.

Small, focused companies are often better equipped to address an un-served market need because they are capable of innovating at a rapid pace. Our small size is one of the reasons that I believe OctigaBay will be successful in introducing a significant innovation to the market.

HPCwire: What do you view as the unmet needs in the high performance computing market?

JOACHIMS: Ultimately, market need in HPC translates to delivered performance on real-world applications. It can be an auto-manufacturer simulating tests for a five-star crash rating, or a computational biologist generating expression sequence tags on new compounds, or one of the national labs working on the grand challenge problems-it all boils down to harnessing technology to deliver the performance required to run the world's most demanding applications. The most significant issue that the HPC market faces today is with the complexity of highly scalable systems. The underlying technology of the computer systems should be transparent; but the fact is that many of these systems require a team of experts to use and manage them.

HPCwire: With that in mind, what do you think the future looks like for high performance computing?

JOACHIMS: From a HPC vendor's perspective, there is definitely a rapidly emerging market for commodity clusters, and with further innovations to simplify usability and manageability, the segment will continue to grow based on simple economics. However, many HPC applications can bring a commodity cluster to it's knees with the demand that they place on the interconnect system. Ultimately, the lack of a solution to that problem leaves wide-body SMPs remaining as the popular choice in the HPC market. Yet SMPs too are plagued with issues; they have difficulty scaling beyond 64 processors, are proprietary and cost-prohibitive. The time is right for a new class of high performance computer to emerge.

HPCwire: What major changes have you witnessed in the high performance computing arena?

JOACHIMS: I would point first to the rapid migration from vector systems to RISC-based systems as the most significant event in the past 15 years - it felt like somebody screamed "student-body-left" and the buying preference of the entire market shifted within a year. The reason for that massive change was simple economics: RISC systems delivered comparable or adequate performance on the vast majority of HPC applications, at about one-tenth the cost. About 5 years prior to that major market shift, we saw the introduction of "affordable supercomputers" which delivered supercomputing performance to a much wider audience. The market shift was not as big, but it was significant and blazed the trail for RISC systems.

Today I believe the prevailing winds of commodity technology acceptance with the early adopters of the market signals another significant shift in buying preference. It's absolutely inevitable that tens, hundreds and thousands of commercially available processors will need to be interconnected together. What's needed is the interconnect speed and other infrastructure elements to make this quest for power successful.

HPCwire: Where do you think further innovations in this space will come from?

JOACHIMS: While the world's renewed interest in high performance computing will encourage the traditional vendors to commit more to their HPC product portfolio, I believe that many of the innovations will come from smaller companies, like OctigaBay. I think there is going to be a whole generation of companies who do not rely on general purpose servers to tackle this target market, but design purpose-built solutions to solve high performance computing user needs.

HPCwire: You commented on the resurgence of interest in supercomputing. What's your view on it?

JOACHIMS: Competition is always good for a market, and Japan's introduction of the world's most powerful computer-the Earth Simulator-provided a catalyst for US policy makers to review investment decisions regarding high performance computing. At the same time, policy makers and national pride aside, it has become increasingly obvious that there is a business need for high performance computing. Whether it's the government's business of providing homeland defense or a pharmaceutical's business of bringing the next drug to market or a manufacturer's business to increase safety in products, ultimately high performance computing is one of the most critical tools they can have in their arsenal. If a car manufacturer can reduce simulation time from five days down to four, that's a real competitive advantage. They need the HPC tools to do that.

HPCwire: How can a young company like OctigaBay develop the HPC products necessary to succeed in this market?

JOACHIMS: As with all companies, it boils down to execution. I've seen a lot of companies come and go in this space; some have had great technology but were not able to execute on their vision. Too many companies become enamored with their own technology, and don't spend enough time concentrating on customer requirements. The first year of OctigaBay was spent talking to a wide variety of potential customers-ranging from the national labs, to life- science companies, to heavy rig manufacturers. The key will be to take the OctigaBay technology embraced by the early adopters like the national labs-who provide some of the best product hardening feedback imaginable-and translate that into broader market acceptance. And, that's where my skill-set in business development will come in. So far, we've had resounding interest from customers and all indicators are that we're heading in the right direction.

HPCwire: When are we going to hear more about OctigaBay's product?

JOACHIMS: You can hear more about us in early November and you can also visit our Booth 2125 at SC2003 in Phoenix, starting November 17th.