Skip navigation

Google Buys 'Server Booster' PeakStream

Google has bought PeakStream, which makes software for high-performance computing.

The Register certainly has a way with a headline: "Google shivs server crowd with PeakStream buy." The news behind the headline: the PeakStream deal could lead to enhanced performance for Google's data centers, allowing its software to take full advantage of the latest chip advances.

PeakStream develops tools that boost the performance of single-threaded applications on multi-core chips. "The PeakStream code looked to benefit the server world at large, making players such as Nvidia, Intel, AMD, Sun Microsystems, Microsoft and IBM the most likely purchasers of the start-up," Ashlee Vance writes. "Now it would seem Google wants to take all the performance gains for itself, perhaps by dabbling with GPGPUs or simply by tuning its in-house software to run well on multi-core chips."

The alarm:clock blog has additional background on PeakStream and its niche: "PeakStream's product is aimed at the high-performance computing market, specifically customers in the oil and gas, defense, and finance industries, as well as in academia. Google says that it will make the products available to their uses, but our guess is that Google will use Peakstream for internal reasons, to make its own machines run better, and to own that secret sauce."


Other industry observers reached a similar conclusion. "I think the reason Google bought this company is pretty straightforward: Google has a voracious appetite for parallel computing power, and that's Peakstream's business," writes Jon Stokes at Ars Technica. "My bet is that the company plans to use Peakstream's software internally."

The official word from the Googleplex: "We believe the PeakStream team's broad technical expertise can help build products and features that will benefit our users," Google said in a statement. "We look forward to providing them with additional resources as they continue developing high performance applications for modern multi-core systems."