Federal Reserve checks out new servers
- By Patricia Daukantas
- Oct 03, 2003
The Federal Reserve System needed more computing power for its check-clearing operations, and the most cost-effective upgrade proposal came from its in-house IT staff.
Dennis Heidlebaugh, a capacity planner with the Federal Reserve IT group in Richmond, Va., used a modeling tool to estimate how much to boost the server capacity at one of the Fed's processing centers. Heidlebaugh spoke yesterday at a Washington conference sponsored by BMC Software Inc. of Houston.
His group was asked to take over systems management for the Fed's check-processing business, which has private-sector competitors. The check application, running on IBM SP servers under AIX, was poised for significant growth and experiencing some performance problems, he said.
In the short term, the IT group used the Predict function of BMC's Patrol for Unix-Perform & Predict to try out combinations of existing equipment and transfer some of the workload to a standby server node. For the long term, however, hardware would have to be upgraded or acquired.
Working with projected check growth, Heidlebaugh's team drew up scenarios for two-hour peak processing periods and modeled them with Perform & Predict. The first set of models predicted that several CPU nodes would become overloaded. Based on the models, Heidlebaugh recommended that the processing center upgrade its hardware to IBM p670 eServers.
That configuration turned out to be less expensive and more expandable than proposals from four outside vendors, he said, partly because fewer IBM DB2 licenses would be needed on the more powerful servers.