Beyond Processing Power

When properly implemented, grid computing can benefit at every level.

When technologies move from the laboratory to the enterprise IT stack, they must make the transition from dancing bear to chorus line. Research projects can claim success if they dance at all, whether or not they dance well, but an enterprise solution must dance consistently—and well enough to go almost unnoticed in the background of a business process.

Grid computing is well on its way to meeting that standard, with significant vendor offerings in database, storage and application development support.

"In three or four years, grid could be the dominant new architecture," said Bob Shimp, Oracle Corp.s vice president of technology marketing. "In large companies and outsourcing firms, it will be the dominant strategy for new capacity."

IBM last year brought the scalability and price/performance of the grid to financial services at discount broker Charles Schwab & Co. Inc.

IBM and Charles Schwab grid-enabled an existing application and deployed it with the open-source Globus Toolkit and Red Hat Inc.s Red Hat Linux on IBMs xSeries 330 hardware. Cycle time for the Charles Schwab wealth management application dropped from more than 4 minutes to about 15 seconds, according to a joint announcement by Charles Schwab and IBM.

IBM offers detailed information and services support to users of its DB2 Universal Database product for integration of database management into the Globus-compatible OGSA-DAI (Open Grid Service Architecture-Data Access and Integration) framework. This enables Simple Object Access Protocol exchanges between an OGSA-DAI user and a corresponding Web service container.

A more off-the-shelf approach to grid-enabled database capability comes from Oracles Database 10g product, which automates the provision of both storage capacity and processing power to multiple databases running in grid environments. Virtualization under Database 10g presents multiple storage devices, or multiple compute servers, to the administrator as single logical units; physical devices can be added or removed from the grid without interrupting database operations.

Database 10g, which became available to customers earlier this month, had previously produced the first Transaction Processing Performance Council TPC-H decision support benchmark results at terabyte scale on Linux, suggesting Database 10gs grid potential.

Grid-based storage offerings have also come forth from major players such as Sun Microsystems Inc. (using technology acquired in late 2002 from Pirus Networks Inc.) and from innovators such as Network Appliance Inc., whose gFiler combines separate devices and NAS (network-attached storage) or SAN (storage area network) installations into a common storage pool with automatic load distribution capabilities.

Further enhancing the illusion of a single system are high-performance interconnection offerings from Sandial Systems Inc., which offers storage network capacity allocation on a per-connection basis to reflect application priorities, and from InfiniCon Systems Inc., which last month announced a partnership with Sun to provide InfiniBand-based grid connectivity at up to 30GB per second.

Grid options

Evolving IT stack virtualizes devices and eases deployment of conventional applications

  • Oracle Database 10g
  • IBM DB2 Universal Database Grid ServicesStorage
  • Network Appliance gFiler
  • Sun N1 Storage ArchitectureInterconnection
  • Sandial Storage Network Backbone
  • InfiniCon InfinIO switchApplication Development Tools
  • IBM Grid Toolbox
  • Oracle JDeveloper 10g
  • HP Adaptive Enterprise
  • Sun Grid Reference Architecture, Grid Engine

Crucial to any IT paradigm shift are application development tools. Oracles forthcoming JDeveloper 10g, for example, complements the abstraction of the grid platform with the abstraction of a declarative programming model that reduces the need for developers to deal directly with procedural task flows among multiple processing nodes.

IBM, through resources including its DeveloperWorks Web site, promotes developer education in the use of tools such as Perl and Perl Commodity Grid Kit for submitting work to grids and managing grid operations.

Hewlett-Packard Co.s development initiatives focus on support for emerging standards and developer education on grid concepts, but HP also promotes access to versions of Globus Toolkit for its platforms.

Finally, Sun supports grid development with platform abstraction and management aids as well as tools derived from its experience in cluster applications.

By putting cost-effective computing power behind the abstraction of the grid, enterprise developers and IT system builders can put more of their effort into unique applications that more closely reflect the needs of the business.