GRID COMPUTING IS NOT MAINSTREAM YET BUT IT WILL BE

The concept of grid computing has been around for a while and has attracted the attention of all the major computer vendors. There is the usual plethora of websites and print publications delivering all the fast-breaking, grid news.

If you're like most of us, you've largely ignored this frenzy. Maybe it's time to start paying attention.

In simple terms, grid computing describes a distributed collection of independent computers that can be assigned to work on different parts of a common problem.

These computers are not centrally managed or administered. They are free to determine when to participate and how much effort to exert.

Grids may be distributed across a department, an enterprise, an industry or the Internet. The advent of web services and service-oriented architecture (SOA) has made grids more practical. These industry standards enable independent, loosely-coupled systems to work together.

For a grid to be effective three elements are needed: a communications protocol, a standard way of describing services, and a format for exchanging information. Web services and SOA standards provide all three via SOAP, WSDL and XML respectively.

Grids are sometimes confused with clusters. A cluster is a tightly integrated and centrally managed set of computers. The computers are usually in close proximity and purchased from a single vendor.

One of the oldest grid computer projects is SETI@home hosted by the University of California at Berkeley. The Search for Extra-Terrestrial Intelligence project analyzes radio transmissions from outer space using desktop computers around the world. Anyone who wants to participate simply downloads the software and installs it as a screensaver. Whenever the computer is idle long enough for the screensaver to activate, the software analyzes data that has been retrieved from central servers.

This example points out a few of the advantages of grids. The first is maximizing resource utilization. The PCs doing the data analysis would otherwise be idle. Thus the SETI project is not taking time or resources away from the user's normal activities.

The second advantage is cost savings. Rather than purchase vast amounts of computer horsepower at great expense, the SETI project simply leverages existing equipment.

Lastly, the grid structure provides the advantage of being highly reliable and resilient. The results are not dependent upon any particular computer or network segment. The analysis continues as long as any computers can access servers at the University of California.

The SETI grid takes advantage of idle desktop computers but there are no restrictions on the types of computers that may be part of a grid. Grids may be composed of servers, clusters, mainframes and/or supercomputers located in multiple data centers. Some of those systems could be dedicated to grid operations while others participate occasionally.

The key idea is to distribute the workload such that resource usage is maximized and response time is minimized.

Grids began in research and educational environments as a kind of poor man's supercomputer. They are now making a serious move into mainstream computing. Here are some examples of the commercial applications of grids:

Molecular analysis for drug discovery by pharmaceutical firms
Portfolio analysis by financial firms
Weather simulations by meteorologists
Collaborative engineering design and modeling by the automobile industry
Special effects by the movie industry
Decision support and data mining by multiple industries

Vendors will often attempt to segment the grid computing market into discreet types using names such as compute grid, information grid, service grid, or intelligent grid. These segments are useful for marketing purposes but in reality they are not mutually exclusive as every grid contains aspects of each type.

A related concept making technology news is "utility computing". Adherents believe that computing resources (processing power, storage, databases, etc.) are becoming commodities similar to electricity and municipal water supplies. Thus, these resources should be deployed so they can be rented or leased as needed. This could be done within an enterprise or supplied by a third party.

Truly useful and reliable utility computing requires considerable maturity of the grid computing model. The management of a diverse and distributed set of computation devices needs better definition and real-world testing.

If you think grid computing may help your business, answer the following questions:

Does your company have compute-intensive applications (e.g. design, modeling, analysis and simulations)?

Do you have computer systems sitting idle while others are over-taxed particularly during peak times such as the end of the quarter or major holiday periods?

Could quality of service and disaster response be improved by creating a distributed environment?

Grid computing is here and ready for enterprise deployment. It's not plug-and-play, however. You'll need to work closely with a knowledgeable vendor and you may need to have some custom software written. With patience and planning, the rewards can be significant.

For more information about grid computing visit the Enterprise Grid Alliance. The EGA is a trade group working to accelerate the deployment of grids in enterprise data centers.

Vin D'Amico is Founder and President of DAMICON, your ADJUNCT CIO™. He is an expert in IT Business Continuity Planning, Network Security Policies, and Freelance Writing focused on white papers, case studies, and handbooks. DAMICON services firms worldwide.

This article appeared in Vin's monthly Virtual Business column for the IndUS Business Journal in July 2006.

To learn more about how DAMICON can help your business, please take a look at our service programs.