CNET también está disponible en español.

Ir a español

Don't show this again

Tech Industry

Commentary: Getting to know 'grid' computing

With companies such as IBM and Platform Computing now touting their capabilities, it's time for executives to get realistic about grid computing's potential--and its shortcomings.

Commentary: Getting to know 'grid' computing
By Forrester Research
Special to CNET News.com
July 21, 2003, 11:30AM PT

By Galen Schreck, Analyst

The initial hype around "grid computing" has faded, but executives still wonder about its application to their companies. With companies such as IBM and Platform Computing now touting their capabilities, it's time for executives to get realistic about grid computing's potential--and its shortcomings.

Despite broad misunderstandings about its real application, companies and users have both been speculating about the application of grid technologies. To date, most grid applications have been custom-built for very specialized tasks, but executives wonder whether they can apply them more broadly to their computing needs.

What is grid computing?
Originally, grid computing applied to massively parallel processing environments built from large clusters of smaller systems. More recently, the development of standard protocols like the Open Grid Services Architecture has encouraged companies to apply the grid ideas to other problems. For example, companies such as IBM have been trying to extend the meaning of grid computing to symmetrical multiprocessing environments to create a highly scalable environment for applications like WebSphere or DB2.

Can grids really solve enterprise business problems?
Yes--in some cases. Today's grid options are most appropriate for a certain class of problems that are easily handled in parallel processing setups, such as sequencing genomes, analyzing stock portfolios, optimizing ASIC designs or searching for extraterrestrial life. But today's grids aren't good for running more mainstream enterprise applications, such as customer relationship management software from Siebel Systems or enterprise resource planning software from SAP.

Is grid computing the same as "organic" IT?
No. Organic IT is a new data center architecture describing the virtualization of the four major infrastructure components: servers, storage, networks and applications. Grid computing is one way that companies can


Related Story

Big Blue says additions to its WebSphere
server software will save customers money
on underutilized hardware.


"virtualize" the server layer of their infrastructures, but it's not the only way. For example, organic server environments might also use virtual machine technology from companies like VMware to run multiple environments concurrently on the same physical hardware, or they could run rapid server-provisioning software from companies like Opsware to reprovision servers on the fly.

What are the different types of grids?
"Grid" has taken on various meanings, and at least two major types have emerged. Compute grids are what you might expect: scalable arrays of processing nodes that spend their time running applications or solving complex problems. Data grids, on the other hand, exist as a sort of network-level file system. For example, IBM has Information Accessibility grid offerings designed to help customers access their disparate data and file resources through common interfaces.

What can grids do out of the box?
Life sciences, financial services and oil exploration companies custom-build many compute-grid applications. These applications, which are easily done in parallel processing, require users to do substantial integration with grid tool kits. Increasingly, however, vertical software makers like Analog Design Automation, Matrix Science and Schlumberger are doing the work themselves, integrating their applications with grid software from companies like Platform Computing.

Are grid products ready for prime-time use?
Yes--when also using rapid provisioning software, many companies can harness idle computing power during off-hours or extended idle periods. Meanwhile, systems companies Hewlett-Packard, IBM and Sun Microsystems are each improving their intelligent workload-management software, allowing companies to expand or contract the resources used by an app in real time.

Which companies lead the grid market?
All the major systems companies, such as HP and IBM, have their internal grid initiatives well under way. In addition, small grid specialists like Avaki, Platform Computing and United Devices have made significant inroads with custom grids for companies in industries like financial services and life sciences.

© 2003, Forrester Research, Inc. All rights reserved. Information is based on best available resources. Opinions reflect judgment at the time and are subject to change.