Automation know-how Outsourcing meets automation Organize to automate Grid on the job Danger in the new data center

When network executives think of grid computing, they often envision the supercomputer-strength computational resources needed for academic or government research projects. That's certainly how IT leaders at human-resources outsourcing and consulting firm Hewitt Associates once saw grid.

About 18 months ago, while working on on-demand computing and other advanced technology initiatives, Dan Kaberon, Hewitt's director of computer resource management, told IBM that no business applications existed for grid. But then he got to thinking about one particularly nasty IT-related business problem the Lincolnshire, Ill., company was coping with, and changed his mind. So after being a vocal naysayer of using a grid for business applications, "it was then up to me to find one," he says, with a laugh.

Working in partnership with Tim Hilgenberg, Hewitt's chief technology strategist for application development, Kaberon created a Linux blade server grid, ported portions of a troublesome mainframe application to it and reduced transaction costs by 90%. In the process, Hewitt demonstrated that grids can solve pressing business needs.

At issue was a mainframe-hosted application that performed complex pension calculations for 5.5 million of Hewitt's customers, many of whom work for multinational corporations that have grown through acquisitions. The pension application, written in Smalltalk running under Customer Information Control System (CICS) and DB2, uses input data such as length of employment, vested percentages, pension payout requirements (which grow complicated after acquisitions and reorganizations), performance of investments and other statistics. It then crunches those numbers to produce estimated monthly pension payments. Users also can perform what-if analysis to model various retirement options.

The application was very popular, but use would fluctuate wildly. Whenever rumors of workforce actions such as mergers, acquisitions or early retirement programs circulated, thousands of employees would go online at the same time to perform pension calculations, Hilgenberg says.

Obviously, Hewitt had no way of knowing when such murmurings would happen. "The application consumed over 1,800 mainframe MIPS and volume could double without warning," Kaberon says. Those big swings of volume "would consume huge amounts of computer power, and, at that time, all of that on a mainframe ... a very expensive place to run something as mathematically intensive as pension calculations," he says.

Whitepapers

• Symantec State of the Data Center Report
• Gene Kim's Practical Steps to Mitigate Virtualization Security Risks
• Monitoring and Managing App Performance
• Monitoring and Managing VoIP
• SNMP Monitoring: One Critical Component to Network Management
• SNMP Monitoring: One Critical Component to Network Management

View more DATA CENTER whitepapers

Webcasts

• Managing the End User Experience-A Real World Example of Success
• Best Practices for Deploying WAN Optimization for Disaster Recovery
• Tips and Techniques for Remote Data Backup and Delivery
• WAN Optimization Landscape
• Oracle Database 11g: Advances in Compression, Real Application Testing and Data Guard
• Solutions to the Toughest IT Challenges in Remote Offices

View more DATA CENTER webcasts

Special Reports

• Data Centers: At the center of the action
• Executive Guide: Data Center Decisions
• Get More From Your WAN
• Executive Guide: Storage Heats Up
• Network World Executive Guide: Perfecting Application Performance Management

View more DATA CENTER special reports