Grid computing: A skeptic's perspective

Spearheaded by such large corporations as IBM Corp., Sun Microsystems, and Oracle, grid computing has become the latest, greatest thing in information systems architecture—at least from the marketing perspective. According to the public relations departments in those and other stakeholder organizations, grid computing is going to solve capacity and performance problems for a myriad of applications. If you need a system that is bigger and faster, grid computing is your solution. But I wonder.

What it is
The basic concept behind grid computing is not really new. The benefits of parallel processing and computing clusters in scientific research are well-documented. The ability to perform billions and billions of calculations using the power of many processors acting in concert is one of those awe-inspiring stories we geeks read about with delight in Scientific American magazine. The success of the SETI at Home project and the University of Pennsylvania's breast cancer research computer grid lend a definite amount of credibility to the overall concept. All of us can understand and appreciate the benefits of this technology.

Grid computing takes the fundamentals of cluster computing and puts an enterprise-friendly face on it. Using the management features of the grid software, the entire computing resources of an enterprise, no matter how spread out and no matter how seemingly incompatible, can be used to tackle a single task. The power of using all of the once-idle CPU cycles connected by an enterprise network is undeniable, but is it practical for everyday, run-of-the-mill operations?

Some of the successful implementations of grid computing technologies can be found on IBM's grid computing Web site. For example, Charles Schwab used IBM's solution to reduce the processing time of a wealth management application from four minutes to about fifteen seconds. There are thirteen other success stories used by IBM to sell us on the benefits of its grid computing technology.

In the case of Sun Microsystems, one of their showcase grid computing success stories is Ford Motor Company, which used the Sun ONE Grid Engine software to implement a collaborative engineering high-performance computing solution for their power train department. The system helped boost productivity in the department by allowing engineers to run thousands of "what-if" problems simultaneously.

What it isn't
While I accept that grid computing or cluster computing is useful for applications that require extraordinary amounts of computational power, my professed skepticism about the technology revolves around grandiose statements that this solution is applicable to the general business enterprise. I am not convinced that super-computational power is required for the accounts payable system or to run your e-mail client. Grid computing seems to have very specific and very limited uses. And while those uses are vital to certain operations and applications, the benefits cannot as easily be transferred to the general enterprise.

As with any system architecture, the benefits of the technology must be weighed against the cost of implementation. The communications system in the computing network, the increased resource management complexity, and the significant cost of the grid application software itself, will all conspire to limit the market for this architecture to large enterprises with specific super-computational needs.

What does the future hold?
My doubt notwithstanding, the real proof of the pudding is what is happening in the developer's universe. In a recent discussion post, I asked the Builder community for examples of projects they are working on, which involve grid computing. I was mostly curious to see what problems grid computing was addressing. As I write this, there has not been one single post showing how grid computing is solving a real-world problem.

Am I missing the boat on this question? Is grid computing actually the next information technology revolution, as claimed by Deb Mielke, or is it merely another system architecture, albeit one with specific uses limited to super-computational problems? As far as I can tell, this technology just doesn't have enough general purpose potential to warrant the current marketing hype. But, maybe, as analysts Anne MacFarland and Mike Kahn from the The Clipper Group Explorer suggest, I need to change my perspective. Let me know. Am I on target or off base on this technology's potential?