Grid computing has been a hot topic for the past two years. but looking at things honestly, grid technology has not yet made much headway in the enterprise. It could be that a perception gap, which we wrote about in a February 2004 special report on grid adoption, still exists.
There does appear to be a mismatch between the grid technologies that are available and enterprise needs. Without question, grid-style architectures have filled the bill in some instances. Where supercomputers once crunched numbers at universities and research facilities, grids are stepping in. And at certain enterprises such as eBay, where extraordinary fault tolerance is required to serve a global 24-hour marketplace, only a grid of interconnected servers across multiple data centers will do. Still, many enterprises dont yet have the needs that can be uniquely satisfied by the grid paradigm.
Further, there are other needs to which IT must devote its attention. Security, for example, is a nearly universal priority. Here, grid has offered no advantage. Adopting grid computing on top of questionably secure platforms or as infrastructure for insecure applications can compound security problems as data and processes spread through more venues and along more connections. These issues also arise in the exploration of Web services; grids and Web services are thus evolving along the same lines in response to the same pressures and are incorporating shared standards as shown in the latest Globus Toolkit. Indeed, Globus Toolkit 4.0, available now, adds support for Web services standards, Security Markup Language and Extensible Access Control Markup Language.
We think the maturity of grid technologies, as exemplified by Globus Toolkit 4.0, will lead to wider adoption of grids. However, it could be that grids within the enterprise will never be mainstream, that their future lies outside the enterprise as the platform for public IT utilities. Vendors are lending their weight to this view. Sun Microsystems this month announced an expansion of its Sun Grid program, in which customers can access computing resources hosted in global data centers for $1 per CPU hour. In addition, the Enterprise Grid Alliance, of which Sun is a member, unveiled its grid Reference Model that could enable standardized deployments of utility computing solutions.
This view coincides with that of author Nicholas G. Carr, noted for his “Does IT Matter?” thesis, who writes in this weeks Free Spectrum that the in-house IT resource will be replaced by an external grid—following the same evolutionary pattern as electricity generation.
But grid-based utilities are not anywhere near widespread adoption. So what do we do until IT becomes like electricity? Forward-looking IT organizations should monitor grid developments but spend resources today on solving problems of application vulnerability and data exposure that no enterprise platform can tolerate. Otherwise, grid computing is not even really an option, let alone a compelling opportunity.
What do you think? Send your comments to eWEEK@ziffdavis.com.
Check out eWEEK.coms for the latest utility computing news, reviews and analysis.