YouTube Facebook LinkedIn Google+ Twitter Xingrss  

Turbo Charged Pharma Computing


By Kevin Davies

Feb. 12, 2007 | ASPEED Software’s Kurt Ziegler has spent two decades helping financial, energy, and other sectors grasp the latest supercomputing phenomena and seamlessly tap grid, cluster, multi-core and other high-performance computing (HPC) environments. But lately, his attention has turned to the biopharma crowd.

Bio•IT World caught up with Ziegler, who is executive vice president, marketing and product management, for his insights into the latest Top500 supercomputer rankings (see “IBM Tops Supercomputing,” Bio•IT World, Dec. 2006/Jan. 2007, p. 35) and the role of HPC in biopharma.

“IBM’s ongoing dominance in the HPC space goes well beyond gigaFLOP’s,” says Ziegler. “It appears IBM’s continued focus on applying technology to solve industry specific problems — while building on previous successes — is paying off.” This is largely because of IBM’s ability to blend technology investments with keen application trend awareness.

Ziegler also sees architectural trends emerging. “Shared memory symmetric multiprocessor system architectures have seen their heyday,” he says. “The new model appears to be multiple independent processors interconnected as a cluster or compactly packaged with a module or frame for economical scalability, footprint, energy, and ease of exploitation.”

Multi-core Movers
IBM’s Blue Gene is an excellent example of such an implementation. Each independent system may be constructed of multi-processors with multiple cores, making it very difficult to classify, but easy to see the performance results.

Another trend is the dramatic growth in applications addressing operational and logistical problems. “This is driving many applications to demand high-performance departmental server, desktop, and even laptop solutions that leverage multi-core and potentially exploit auxiliary processing capabilities, which are currently only available to gaming and graphical products.”

Ziegler says the emergence of multi-core processing has the potential of accelerating life sciences breakthroughs by providing increased personal and enterprise computing horsepower, energy efficiency and price-performance. “Multi-core is leading the way to new programming,” he says. “It promises to become the change agent to dealing with more complex problems, which until now were tabled because of computational time constraints.”

Multi-core solutions offer additional capacity at relatively low costs. “The ramifications go far beyond sheer performance, reduced energy, and small footprint: Multi-core can significantly reduce the operational complexity of running high performance applications giving researchers and analysts more flexibility and control.”

“There’s IT, and then there’s what the scientist wants on his desktop. Economics is one reason. Given the availability of dual cores and multi cores on the desktop, [scientists] can suddenly do models [they] never could do before, in less time, and with better precision. Pharma scientists were using grid technology, but can now use 4- and 8-way machines at their desk.”

An analysis that might take four hours on a single processor/single core system could be halved with a dual processor, says Ziegler. But on a dual-processor, quad core desktop system, the same analysis could be slashed to a mere 30 minutes. “Researchers could be very productive on their desktop system without having to schedule or share capacity until a very time consuming run, which would be distributed to servers, clusters, or grids,” says Ziegler.

ASPEED is working closely with Intel and ISVs on quad-core technology advantages. Benchmark testing on the new Intel XEON Quad Core 5300 processor using ASPEED’s ACCELLERANT showed very dramatic performance improvements, up to 12 times faster (on 8 cores). “The new Intel multi-core systems have added other improvements such as cache handling, bus architectures, and pipelining, which can all be taken advantage of when the application is running in parallel,” says Ziegler.

Applications
Several Top500-ranked supercomputers including Blue Gene are running life sciences applications, including the Computational Biology Research Center of The National Institute of Advanced Industrial Science and Technology for 3-D protein structure prediction, and Gene Network Sciences with its cardiac modeling software, incorporating data on a drug’s effect on a cardiac ion channel to measure arrhythmia.

ASPEED is working with independent software vendors such as GloboMax’s NONMEM (used by the FDA to get drug approvals) and Pharsight’s Trial Simulator. “Early users are seeing reductions between 25-75 percent in their analysis time enabling them to conduct more “what ifs” and exercise larger models in less time,” says Ziegler.

ASPEED’s solution “accelerates applications,” says Ziegler. “It’s software that’s an additive, substantially improving properties of the application.”

Email Kevin Davies.

Subscribe to Bio-IT World  magazine.

 

Click here to login and leave a comment.  

0 Comments

Add Comment

Text Only 2000 character limit

Page 1 of 1

For reprints and/or copyright permission, please contact  Jay Mulhern, (781) 972-1359, jmulhern@healthtech.com.