Microsoft Tests Speedy New HPC Server

SANTA CLARA, Calif.—A lot of people don't know this, but Microsoft, which got rich producing PC operating systems and applications, has been investing heavily in research and development in high-performance computing for about the last eight years.

In fact, the world's largest software company currently employs hundreds of engineers whose sole job it is to conceptualize and develop new products for the burgeoning HPC market, which research company IDC reports is the fast-growing sector in IT at 19.5 percent per year.

Microsoft's No. 1 product in this arena is its Windows HPC Server 2008, now in beta trials among selected customers. It is expected to be finalized sometime in the summer of 2008, Microsoft General Manager of Worldwide Partner Evangelism Eddie Amos told attendees here at the Platform Global Conference held May 19 to 21 at the Hyatt Regency.

HPC Server 2008 will succeed the company's first HPC server, Windows Compute Cluster Server 2003, which was released in 2005 but didn't make a large dent in the market.

"We're working closely as a key partner with Platform Computing to come up with excellent HPC products," Amos said. "You'll see more along these lines later this year."

The new 2008 server runs on commodity hardware and is 30 percent more efficient in use of power than the 2003 edition, Amos said.

Platform Computing, a specialized HPC management software producer that hosted the conference, gets along well with a number of companies that are dyed-in-the-wool enemies in a lot of other sales areas. The Canadian company has close partnerships with not only Microsoft, but also IBM, Dell, Intel, Hewlett-Packard, Red Hat and SAS Institute.

Click here to read about why the field of high-performance computing is growing rapidly.

In Linpack standardized benchmark testing in Europe in April, Microsoft reported that the HPC Server 2008 delivered about 18 teraflops' I/O performance—pretty speedy in anybody's book.

That particular test had HPC Server 2008 running on 2,048 processors to deliver over 75 percent of the theoretical peak performance for a single application, a company spokesperson said.

"We employed the unmodified HPL code to solve a linear system with 681,984 unknowns in a little over 3 hours at an average speed of more than 18,000,000,000,000 floating point operations per second," the spokesperson said. "The program had a total memory footprint of almost 4TB.

"The cluster consisted of 256 compute nodes equipped with two new quad-core Intel Xeon E5450 processors running at 3GHz clock speed each. All nodes are connected with each other by a fast InfiniBand fabric with a 288-port switch from Cisco Systems. The cluster nodes were built by Fujitsu Siemens," he said.

Amos said trials are ongoing in the HPC Server 2008 group and that Microsoft will continue to ramp up its R&D in the high-performance computing space.