Long discussed, grid computing is alive and kicking and coming to a data center near you. And sooner than you might think.
The mainstream vendors that attended the mostly academic GlobusWorld conference last week in Boston were bullish on grid, saying they have customers piloting grids and looking at significant implementations for 2006.
IBM, HP, Sun and Intel attended GlobusWorld as newly minted members of the Globus Consortium. Formally announced late last month, the Consortium is one of two Globus community efforts to commercialize the widely embraced Globus open source grid technology. The other is the recently formed Univa , a Red Hat-like company that is coming out with a commercial distribution of the open source code.
Both efforts are meant to further development and adoption of the Globus Toolkit, the open source middleware that can be used to virtualize computing environments.
Why grid now? The time is right. By many accounts, average system utilization across organizations is 15% to 20%, while obviously the ideal would be about 80%. And some 20% of IS budgets goes to operations today, marginally less than the 25% that goes to capital investments.
We’ve created large, underutilized, complex environments that are costly to maintain. So there’s a huge need to do this better, and the prevailing thinking seems to be that grid is the answer, even though the big vendors are wrapping grid in terms such as utility, adaptive and on demand.
As luck would have it, the arrival of solid grid technology in the Globus Toolkit roughly corresponds with the adoption of service-oriented application architectures. The grid faithful say these service-oriented applications demand grid-powered service-oriented infrastructures.
Grid won’t arrive tomorrow, but it adds up to a sensible image of the future we can build toward. The main question is whether the vendors can see enough economic gain to play sensibly together and avoid interoperability problems while still advancing their individual causes.
Kenneth King, IBM’s vice president of grid computing, adds it up this way: Cooperate on standards, compete on implementation. “We aim to provide the best software technology to virtualize computing environments and enable grid,” he says.
Part of the problem today is there are too many standards organizations, but the new Globus Consortium will help, as will the creation of Univa. And King says he is seeing more meaningful interaction between standards bodies.
Suddenly, grid seems to be gelling, and we can expect real advances this year and next.
