February 15, 2012

Grid Engine Diagnostic

Robert Gelber

When Sun Microsystems was acquired by Oracle in January 2010, the outlook for much of Sun’s HPC technology was less than rosy. In particular, support for Grid Engine, a distributed resource management system developed by Sun, faltered. In December 2010, Oracle revealed plans to commercialize Grid Engine, in which it ceased to be an open source product.

After the Oracle announcement, one of the major problems Grid Engine users were faced with was having to pay a commercial license for something that was previously free. I like to equate it to paying for Windows service packs. How would I feel if Microsoft charged me for every regular patch and update they provided?

Last January, Austin-based Univa, with the help of former Sun developers, announced that they would provide a directly competing commercial offering for Grid Engine, as well as open source support. In fact, Univa wanted users to be sure enough of their support that they actually proclaimed “We live and breathe Grid Engine” right on their website.

As the technology has traveled this bumpy road for the past couple of years, a lot of HPC’ers have wondered if the technology is still relevant and viable, and if so, which version of Grid Engine should they be using. In Bio-IT World this week, BioTeam co-founder Chris Dagdigian wrote an article set out to answer those questions.

When Dagdigian got involved in a number cloud projects, he began to wonder if Grid Engine’s resource management capabilities had been made obsolete by the overarching cloud computing frameworks. As it turned out, users needed the technology to support legacy research environments and to do the kind of scheduling and resource allocation that is Grid Engine’s forte.

As far as the viability issue, dual-vendor support from Oracle and Univa gave Dagdigian confidence in Grid Engine’s long-term prospects. One might even say that the competition introduced by Univa may have been one of the best things that happened to Grid Engine. There are also two different open source projects supporting the technology: Open Grid Scheduler and Son of Grid Engine.

That said, Dagdigian doesn’t expect Grid Engine to survive under the Oracle regime much longer since it’s not a big revenue generator for the company. His recommendation for users needing a commercial product is to acquire Univa’s offering. For people who want to stay on the open source path, Dagdigian is pointing people to the Open Grid Scheduler project since it now has the option of commercial support.

So the answer to Grid Engine’s relevancy is, to some at least, still yes. Of course, the next question is how long can it last? Will it be able to adapt as HPC and its underlying technologies evolve? Ultimately, the community will determine the technology’s viability and vendor interest. As James Earl Jones might have said: “If you use it, support will come.”

Full story at Bio-IT World

Share This