Microsoft today announced that the latest version of its Windows HPC supercomputing stack has been sent to manufacturing as the software giant continues to advance on a broad front at the very high end of computing capabilities.
Microsoft has been involved in the high-performance computing space for a few years now and the HPC Server 2008 release comes as the firm is beginning to build a significant customer reference list. Although it won’t confirm exact customer numbers or revenue, it does say that it now has “hundreds” of customers. Banner names among them include the Ferrari F1 motor racing team, Lloyds TSB, Boeing and Procter & Gamble, plus several big academic sites. Just as important, Microsoft seems to be stacking up ISV support from the rarefied HPC application world including Mathematica, Schlumberger and ESI.
Throughout, Microsoft has been telling a characteristic story about low cost, standardised development and simplification to lure the HPC community away from Linux.
“We’re aimed at accelerating productivity, innovation and gaining better insight into data,” said Kyril Faenov, Microsoft director of HPC. “We want to reduce complexity for end users and administrators, and we consistently hear that performance is something hardware can deliver but the ability to use that is lagging behind.” Lloyds TSB’s capital markets team set up a 100-node cluster in half a day, Faenov added, whereas “in the past it would have been weeks or months”.
In another characteristic manoeuvre, Microsoft is also touting the virtues of pluralising the technology and putting supercomputing at the disposal of anybody who needs it, from financial services analytics to engineering design. Faenov conceded that he would like to see more ISV support in life sciences where the existence of “lots of little software applications” made it hard to cover bases, but added that gaining traction with in-house developers through HPC support in Visual Studio and .Net tools was just as important.
Further out, Faenov said that cloud-based HPC capabilities could appeal to companies that needed ad hoc access to remote HPC resources. “A small hedge fund might have a small cluster but could maybe do something in the cloud whenever they needed advanced analytics,” he argued, while acknowledging that pricing, security and other issues will need to be ironed out and the market is currently “nascent”.