Enterprise Unix Roundup: Supercomputing Transformed
|Main||In Other News||Elsewhere in the Corral||Tips of the Trade|
Much of the news to hit our RSS readers this week centered around supercomputing: the show, the list, and the buzzword. We've been following supercomputing for a while now, and we've watched the show that bears its name evolve from being aimed at a largely academic and scientific audience to being a major event for end-user enterprises.
This isn't surprising, as the market, which grew 30 percent last year, is expected to reach $8.7 billion in 2006, according to IDC.
At this week's Supercomputing show (known by the catchy moniker, SC 05) in Seattle, Linux was king and Microsoft announced its plans to sell some supercomputing bling-bling, which is as sure a sign as any that supercomputing isn't what it used to be. Sure, it's faster and more powerful, but it's not contained solely in the exotic realms of academia. Redmond taking interest is perhaps the greatest litmus test that something has gone mainstream.
We recapped the list and the show earlier this week, and, as noted previously, saw a resounding theme of standardization. Linux on x86 appears to be the combination of choice, and HP and IBM collectively consumed 77.6 percent of all slots. No other manufacturer was able to capture more than 7 percent in any category.
The players traditionally associated with supercomputing (e.g., Cray, SGI, and NEC) and some surprises (e.g., Dell and Apple) shared this 7 percent.
Microsoft's entrance in the game isn't terribly surprising. End-user enterprises are taking more of an interest in supercomputing, sometimes under the label of grid computing; sometimes calling it utility computing. Microsoft (who, remember, is not an innovator) will take an approach to supercomputing similar to the tact it takes in other areas: It will provide a complete Windows platform for HPC to help customers, who otherwise wouldn't be in the market for a supercomputing solution, to be up and running quicker.
With Linux running on 72 percent of systems that made the list, a Windows' solution may not be as well received as Microsoft would hope. Linux has earned a reputation as being particularly well-suited for clustering, and in an environment like HPC, is plug-and-play an attribute enterprises would want?
We're inclined to think not, and briefings we did in the past few weeks would support this. Cray and SGI are two companies that have placed their future in the hands of Tux. The servers and solutions they are bringing to market, are Linux-based, not proprietary.
One vendor who is simplifying the complexities of supercomputing is Interactive Supercomputing, a Waltham, Mass.-based startup spun from an MIT lab. The company launched Star-P, and interactive parallel computing platform that enables "engineers and scientists to go from prototype to production using familiar tools," Interactive Supercomputing Vice President of Marketing Ilya Mirman told ServerWatch. With Star-P, end users code algorithms and models on their desktops using MathWork's MATLAB and then run them on SGI systems, thus eliminating the need for developers to convert the models to applications in C, Fortran, or MPI a conversion that can take months.
Another interesting announcement, whose timing we suspect was not entirely coincidental with the show, came from Sun. The vendor this week unveiled details about the processor formerly known as Niagara (see details below). Although a mere four SPARC systems made the list this week, if we were gamblers, we'd wager Solaris 10 on the new UltraSPARC T1 CoolThreads will fare better on next November's list.
Ed Note: Enterprise Unix Roundup will not publish next week due to the Thanksgiving holiday. We will return with two weeks worth of news and analysis on December 1.