Review of VA's Revised Workload Study for Its Austin, Texas, Data Processing Facility
FGMSD-80-44: Published: May 7, 1980. Publicly Released: May 22, 1980.
- Full Report:
In 1978, GAO reviewed a Veterans Administration (VA) proposal to acquire an IBM 370/168 multiprocessor system to handle increased requirements for it Beneficiary Identification and Records Locator System (BIRLS). The major cause of the increased requirements was the installation of a new Target System which would use information in the BIRLS master database. The Target System is being completed in phases. As more regional offices are provided with on-line capabilities, the workload increases, especially during the prime-day shift. GAO felt that an IBM 370/158 would be sufficient and proposed austerity actions to eliminate projected peak-hour shortfall. VA officials at first agreed with the GAO position, but later determined that the IBM 370/158 would not be sufficient to meet its needs. In 1979, Congress instructed VA to study the problem further. GAO reviewed the VA study which recommended that VA be authorized to use the IBM 370/168 computer rather than the IBM 370/158 recommended by GAO. The study contained new workload projections, and GAO was asked to reevaluate its earlier recommendation in light of these projections.
The review showed that the workload has increased substantially and that additional capacity is needed. However, the workload projections which VA has prepared have varied widely. Because of these variations, GAO could not estimate with any confidence what size computer should be installed at the Austin computer center. Although VA was unable to explain why the projected workload has increased so rapidly, GAO found two factors that may have contributed to the problem. First, operator carelessness at the terminals, design problems, or improper operating procedures rather than changing requirements could be contributing to the wide variations in workload projections. Secondly, VA has not adequately analyzed the impact of new administrative procedures. GAO did not observe any efforts on the part of VA to validate the projection method used or to identify the causes of the increases. No effort has been made to reduce the peak hourly rate. Two alternatives for meeting the projected workload are for Congress to (1) require, as a prerequisite to acquiring a computer, that VA develop an accurate and valid workload projection, determine the cost effectiveness of implementing the changes recommended by GAO in 1979, and reduce the peak hourly workload by scheduling nonpriority transactions at other times; or (2) if urgency is an overriding consideration acquire a computer with a capacity greater than that of the IBM 370/158, contingent upon a commitment by VA to fulfill the above requirements.