HIGH-PERFORMANCE COMPUTING · The 800-lb. gorilla of desktop computing plans aggressive push into life science clusters

BY SALVATORE SALAMONE 

February 2, 2005 | Microsoft will soon launch a major effort into the high-performance computing (HPC) arena that combines new HPC cluster software and collaboration with Microsoft partners to develop life science-specific computing solutions.

The main HPC push from Microsoft will come in the second half of this year when it releases MS Windows Server 2003, Compute Cluster Edition (CCE). Microsoft says CCE will be a complete Windows-based HPC solution that addresses cluster setup and management, application develop-ment, and job scheduling.

The idea is to offer server software optimized for HPC clusters and to tap other Microsoft systems management and development tools to simplify management and application development for clusters. For example, Microsoft plans to offer compilers and a parallel programming debugger with CCE, as well as features to deal with message-passing techniques commonly used to move data in a cluster.


Gauging User Interest
While Linux clusters increasingly dominate life science computation, many believe there is an opportunity for Microsoft to play a role in HPC. Two areas seem particularly ripe for a Microsoft entrée into HPC:

• Many Windows applications, which today run on a single computer, could run much faster if they were migrated to a Windows cluster.
• A large body of custom software that runs on Unix SMP machines could just as easily be ported to Windows as Linux if needed tools become available.

Still, the question remains: Will life science researchers take Microsoft's efforts in HPC seriously? A few have already bet on Microsoft, and their reasons provide practical insight into Microsoft's strengths in this market.

For example, the Cornell Theory Center (CTC) and its Computational Biology Service Unit chose Windows for its HPC needs for several reasons. "[I] wanted someone else's butt on the line if software didn't work," says David Lifka, chief technology officer of the CTC. "We had a [level of support] with proprietary Unix." When moving off these Unix systems, "we didn't want to have to go to a news group or pay extra for this [support]," he says.

"Biologists want to do biology. The main thing is to get the operating system out of the way." David Lifka, Cornell Theory Center

Lifka insists that as long as the computing capacity requirements are met, scientists don't much care about the underlying operating system. "Biologists want to do biology," he says. "The main thing is to get the operating system out of the way."

Easy-to-use development tools are another attraction, Lifka says. "Developing applications with [Visual Studio .NET] takes a very small amount of time compared to other approaches," such as Java and CGI.


Partners Are Key 
Part of Microsoft's strategy is to do what it always does — leverage its strength with independent software vendors (ISVs). "Our goal is to build a partner ecosystem around Windows Server HPC," says Greg Rankich, senior product manager in the Windows server group at Microsoft. Moreover, the software giant is specifically targeting the life sciences, manufacturing, and energy/oil/gas industries in its push into HPC.

"We looked at the cluster-ready applications," Rankich says. "Life sciences came out number one." He points to BLAST, Gaussian, and other informatics algorithms as the types of applications that fit into this category. Rankich also cites similarities with the manufacturing sector in which many computer-aided design (CAD) applications traditionally ran on Unix SMP servers and workstations and are increasingly being ported to Windows and clusters.

In the process of building its partner ecosystem, Microsoft has already been working with a number of life science software vendors. For instance, it has worked on a project with molecular discovery software vendor Optive Research (recently acquired by Tripos).

"Our collaboration with Optive Research [helped] Microsoft better understand the unique HPC requirements facing discovery scientists at many of the world's leading pharmaceutical companies," says Kyril Faenov, director of HPC at Microsoft. Faenov says those interactions and collaborations will help ensure that Windows Server 2003 CCE "meets the needs of the life science community."

Optive Research worked with Microsoft on two fronts: redesigning its software on the .NET platform, and making the software easier to use. "For the lab scientist, [a Windows-based application] helps with the ease of use," says Bryan Koontz, former CEO of Optive Research and now vice president for discovery informatics at Tripos.

Koontz notes that when the collaboration started, one question was: Could Windows versions of applications achieve performance parity with Unix or Linux?

For Microsoft, the strategy is simple. If it can eliminate the performance issue, it can sell people on other things such as total cost of ownership, application development support, the wealth of Windows Server systems management tools, and integration with other Microsoft products.

Many organizations use Microsoft's Active Directory. HPC applications running on clusters could take advantage of the user identification and security features within Active Directory in a number of ways. For instance, user access rights might be used to grant or deny access to computing resources or to give different users or groups higher priority to the resources. 






PHOTO BY KEVIN STEARNS/CORNELL UNIVERSITY