Opportunity for Dell as supercomputing costs fall
SAN FRANCISCO |
SAN FRANCISCO (Reuters) - Supercomputing, once the province of a select few, is fast becoming a key business tool, a trend Dell Inc (DELL.O) hopes to capitalize on.
Long used in cutting-edge research, high performance computing (HPC) has become far more affordable in recent years and has expanded into more enterprises, from drug development to automakers simulating car crashes and food companies looking for the best way to package a snack.
The rise of so-called cluster systems -- based on Intel x86 processors -- have helped make HPC 10 times cheaper than it was earlier in the decade, analysts say. The overall market was worth roughly $10 billion in 2008, according to IDC.
Cluster systems are comprised of a series of nodes -- or server computers -- sometimes hundreds of them, that are designed to tackle enormously complex problems.
Although still trailing larger rivals Hewlett-Packard (HPQ.N) and IBM (IBM.N), Dell has been gaining ground in HPC, by selling lower-cost and mid-range systems.
"If you've kind of followed the whole supercomputing space, you've seen a real transition over the last 10 or 15 years to clusters of microprocessor-driven computers," said Chief Executive Michael Dell.
Dell's CEO called HPC a "key vertical" for the company, particularly in its public sector business.
"There are all these grand challenges that are out there... and as we get this exponential increasing computational power, that unlocks the ability to go solve these problems at a faster and faster rate."
At a small gathering at Lawrence Livermore National Laboratory in California last week, Dell executives met with lab officials to talk shop as they marked the installation of a new Dell supercomputer named "Coastal".
Dell -- along with partners including Intel (INTC.O) and Cisco Systems (CSCO.O) -- is also working with the lab to develop an advanced supercomputing test bed system.
Lawrence Livermore, which conducts national security research for the U.S. government, is home to some of the world's fastest computers, including IBM's BlueGene/L, which held the top spot for several years.
TACKLING COMPLEX TASKS
Supercomputers have long been used in fields such as oil exploration, weapons research and genomics.
News about HPC often revolves around the fastest, most expensive systems. The top 500 list, compiled twice a year, provides both bragging rights and a snapshot of how far the industry has progressed.
IBM's Roadrunner, at Los Alamos National Laboratory in New Mexico, is the current champion.
IBM and Hewlett-Packard dominate the top 500 list, together accounting for around 80 percent of the supercomputers on the tally. Machines at the top of the list can costs upward of $200 million, analysts say.
Dell is not a major player in high-end systems, said IDC analyst Earl Joseph. But he said Dell has done well at the lower end, and has made steady gains in the middle as well. An average system costs around $350,000-$400,000, he said.
"The very high-end is not real profitable. But in the middle of the market where Dell plays more, and at the lower end it's real profitable."
Joseph estimates HPC accounts for roughly 20 percent of the server sales combined for IBM, HP and Dell.
Dr. Mark Seager, head of advanced computing technology at Lawrence Livermore, said many companies are now using supercomputers for predictive simulation, saving costs by cutting down on the number of physical experiments.
"Every year the incremental progress we can make is 10 times what we are able to do in the preceding decade. There's stunning breakthroughs that are coming about."
Forrest Norrod, general manager of Dell's data center solutions division, likened the challenge to an arms race.
"There's an insatiable demand for compute power out there and we don't see it stopping or slowing down because it enables new applications," he said.
(Reporting by Gabriel Madway; Editing by Tim Dobbyn)
- Tweet this
- Share this
- Digg this