Original URL: http://www.theregister.co.uk/2009/09/28/microsoft_buys_isc_carcass/

Microsoft munches super startup carcass

Can you say Visual Studio for HPC?

By Timothy Prickett Morgan

Posted in HPC, 28th September 2009 04:15 GMT

Microsoft has eaten the carcass of Interactive Supercomputing, a creator of HPC development tools for parallelizing applications.

That's yet another promising supercomputing startup that has succumbed to the economic Meltdown and the difficulty of finding capital to fund its expansion. Redmond has acquired the company's the intellectual property and key employees.

Kyril Faenov, general manager of high performance and parallel computing technologies within Microsoft's Windows Server Division, made the announcement of the ISC acquisition on the division's blog. The financial details of the asset acquisition were not disclosed, but it is clear that Microsoft is not buying the whole company and it's not assuming whatever debts and obligations the privately held ISC had.

"This move represents our ongoing commitment to parallel computing and high performance computing (HPC) and will bring together complementary technologies that will help simplify the complexity and difficulty of expressing problems that can be parallelized," wrote Faenov in the blog.

"ISC's products and technology enable faster prototyping, iteration, and deployment of large-scale parallel solutions, which is well aligned with our vision of making high performance computing and parallel computing easier, both on the desktop and in the cluster."

Microsoft has been on a mission for the past two years to become a player in the HPC space, and it seems poised to grow its share of the HPC market by getting a hot at greenfield personal, departmental, and midrange supercomputers among researchers who do not have experience with Unix or Linux systems, who have for the most part work with Windows workstations, and who would prefer to stick with a Windows platform as they grow their applications.

Windows HPC Server 2008, Microsoft's second real pass at delivering a parallel clustering stack for Windows boxes, is an attempt to not only mimic the early success that the Linux platform had a decade ago with Beowulf clustering - and thereby putting Linux on the map as far as corporations and governments were concerned - but also to take away some of the HPC thunder of the Linux platform, which has usurped the dominant position in HPC that Unix once had.

ISC created a programming environment called Star-P, which allowed researchers to use familiar software packages such as MatLab or programming languages such as Python to develop their applications on a desktop machine and then have it automatically parallelized to run on a cluster of boxes.

Old Redmond pal

ISC was founded in 2004 by Cornelius Peterson, a serial entrepreneur who sought to crack the parallel programming problem based on research done at Massachusetts Institute of Technology. Previously, Peterson had tackled system-on-a-chip designs with a company called NetSilicon, which was founded in 1984 and went public in the 1989.

In January 2007, the Waltham, Massachusetts, company brought in Bill Blake as the company's chief executive officer, bumping Peterson up to executive chairman. Blake was senior vice president of development at data warehousing appliance maker Netezza most recently, and before that, he was vice president of Compaq's HPC unit (really Digital AlphaServer stuff with some x86 clustering thrown in) prior to that.

Whatever plans that ISC had for growth, the venture capitalists who pumped a reported $18m into ISC since 2005 - Ascent Venture Partners, CommonAngels, Flagship Ventures, Fletcher Spaght Associates, and Rock Maple Ventures - appeared to run out of patience as ISC ran out of money.

The fit with Microsoft is good because the two companies were already working together. Last September, as Microsoft was getting Windows HPC Server 2008 out the door, ISC was demonstrating a preview version of Star-P for Windows HPC Server, and the company said it planned to get the code out the door by the end of the year. Star-P was originally developed to deploy applications on Linux-based clusters.

Faenov said in the Microsoft blog announcement that Blake would be taking a job at Microsoft's New England Research and Development Center in Cambridge, outside of Boston, and would be "bringing over a team of industry leading experts on parallel and high performance computing that will join the Microsoft team."

Microsoft has already begun planning for the integration of ISC code into future versions of Microsoft products, and it says it will continue to maintain the current release of Star-P. If you look at the FAQ put out by Microsoft concerning the acquisition, which is here, you will discover that Star-P version 2.9 - which was released into a controlled beta a few months ago - is not going to be launched as a product and that there will be no releases beyond the current Star-P version 2.8, which came out earlier this year.

<p.Moreover, Microsoft is no longer selling new licenses the Star-P tools, but customers who have active maintenance have the right to upgrade to version 2.8. Microsoft will support Star-P version 2.8 until the end of existing contracts or through December 31, 2010 (whichever one is the longer term).

The most obvious thing for Microsoft to do is to rejigger Star-P as a plug in for its Visual Studio development tools and plunk the parallelization of those applications and their runtime out there on Windows HPC Server 2008. Microsoft said that it will lay out its plans for weaving the Star-P assets into its products and when this will happen "over the coming months." ®