Enterprises are constantly looking for ways to reduce their IT expenses, and hardware purchases represent a significant portion of those costs. “Companies want to avoid buying faster systems year after year and grid computing has that potential,” said John F. Andrews, chief operating officer at market research firm Evans Data Corp.
Grid computing is based on the notion that enterprises are under-utilizing their existing systems. Traditionally, companies bought computers to support specific applications, such as Customer Relationship Management (CRM), Enterprise Resource Planning (ERP), e-commerce, or a company Web site. In some cases, these applications have processing peaks and valleys. As a result, systems can operate at as little as five percent of their capacity during off-peak periods.
Grid computing tries to take advantage of those lulls by moving processing chores from overburdened to under-utilized systems. Another plus is grid systems are designed to work with inexpensive Windows, Linux or Unix computers rather than the costly high-end servers that have traditionally supported business applications.
Grid Computing Is Really Super
The grid concept is new in the corporate environment, but it has been in use for decades with supercomputers. Research organizations, universities, and government agencies rely on a time-sharing model when they lease computing time from one another on an as needed basis.
Vendors have tweaked this model and envision departments in companies sharing the computers that support their applications. “IBM and Oracle have been pushing the grid concept most aggressively,” said Joshua Greenbaum, a principal at Enterprise Applications Consulting. IBM views grid computing as a way to enliven sales of its high end systems and Oracle sees the technology as a way to boost revenue from its database management systems. Hardware suppliers Hewlett-Packard and Sun Microsystems have joined start ups Datasynapse, GridXpert, Platform Computing and United Devices in delivering grid systems.
The technology is gaining traction in select market segments. “Pharmaceutical companies are using grid computing to try and identify new drug treatments,” noted Carl Claunch, a research vice president at Gartner Group. In addition, the hardware option is being used for seismic processing by oil and gas companies, risk analysis for financial enterprise, product design by manufacturers, and simulation by aerospace companies.
Aetna Life Insurance, Chicago Stock Exchange Holdings, Ohio Savings Bank, and the U.S. Geologic Survey are a few of the organizations that have adopted grid computing.
What Exactly Are We Talking About?
While the concept has been garnering interest, it still faces a few hurdles, starting with a clear definition. “There are vendors using the term grid when in reality they are only offering clustering capabilities,” Gartner Group’s Claunch told TechNewsWorld.
Since the middle 1980s, enterprises have divided processing chores up among servers sitting in one data center. Grid takes this concept one step further and relies on systems that sit in multiple locations.
Since this is not how software was designed to run, a lot of infrastructure has to be added to existing systems in order for grid to work. Applications have to rewritten. Since grid computing means that a company is chopping up its computing tasks, software has to be written to divide tasks into small pieces and put them back together. Another issue is currently grid computing isn’t good at trying to decide where or when to carry out a task and may not benefit applications where one action is dependent upon the result of a previous one, which is common in many ERP and CRM systems.
As grid computing raises such issues, firms can spend a lot of time on systems integration and application development. “Because grid is new, there aren’t a lot of companies or individuals who have experience in this area,” Enterprise Applications Consulting’s Greenbaum told TechNewsWorld.
With information running on different machines, there is also a need for new systems management tools. Vendors have been working to add workload management features, which help move information among computers, and failover response functions, which make sure that important transactions are completed. “The management functions found with grid systems are immature at the moment,” said Gartner Group’s Claunch.
Extensions are being made to system management tools, such as Computer Associates’ Unicenter and IBM’s Tivoli, to accommodate a grid environment. Because grid computing is still young, many of the tools are based on proprietary technology. The Global Grid Forum has been working on the Open Grid Service Architecture, a set of specifications to make it easier for transaction-focused enterprise applications to take advantage of grid computing.
Working Its Way Into the Data Center
Until those items are in place, the notion of companies capturing unused desktop or laptop computing cycles is more of a vision than a reality at this juncture. Many of the systems redistributed processing cycles are currently found with servers in data centers.
Technical barriers may not be the technology’s most significant challenge. Platform Computing interviewed 50 companies and 89 percent of respondents reported that organizational politics were the biggest roadblock to grid deployments. Figuring out how many computing cycles are available could be difficult and requires that the IT department collect performance data on all systems.
Corporations tend to operate in standalone fashion with departments responsible for specific applications. In this consolidating environment, new organizational charts have to be developed. “Someone has to be in charge of determining when to take cycles from one machine or another,” noted Enterprise Applications Consulting’s Greenbaum. Some managers may not want to let go and share their resources because they feel they’ll lose control, lose budget and lose priority over their work.
While grid has a lot of potential long term, it will probably be used only in select instances in coming months. “We expect only a small percentage of customers will move their applications to grid computing during the next 12 to 18 months,” concluded Gartner Group’s Claunch.