MAINFRAME SOFTWARE INNOVATIONS
David Cole - Lead Consultant
Client: Military and commercial aircraft division of a large aerospace contractor
The main MRP "Cost" program was running about 2 hours per night processing a mid-sized database. After the client added a major new aircraft to MRP, the database and processing load increased to twenty times the prior load, resulting in run-times of over 40 hours for this one program. Other programs were affected, but only the main "Cost" program was large enough to be affected so severely. Through a variety of improvements to the logic of this program, a version was created which accomplished exactly the same results, but eliminated 90% of the clock time and 95% of the CPU time, thus reducing run-times to about 4 hours per run.
Client: Military aircraft division of a large diversified technology company
The "Roll-Up Loop Check" process involved a jobstream which was executed 22 times in a row (once for each "level" in the order bill-of-material). MSI was asked to add some new functionality to this process and to reduce the run-time if possible. The main program was rewritten utilizing several simple but seldom-used data structures to avoid the need for re-execution, as well as helping to accomplish the new functionality. The result was a jobstream containing just one program, which was executed once instead of 22 times. The reduction in run-time and CPU time was in excess of 75%.
Another project for this client involved a daily user-submitted SAS program that matched two multi-reel tape cartridge datasets. It was running over 5 hours each day, preventing other users from accessing these important datasets. In less than 2 hours the function was rewritten using Dyna-Match, reducing run time to about 12 minutes each day.
Client: Global Information Technology provider
Runtime of a program which performed "part number conversion edits" was reduced from 780 minutes to 16 minutes by converting table handling logic to use of a VSAM file.
Client: Commercial manufacturing company
While working on an unrelated task, we became aware that the batch runtime of a large manufacturing system had been steadily increasing due to the increased numbers of parts and orders being processed as the company's business grew. The available batch window was full and expanding that window could only be done at the expense of reducing online availability of the system, which was not acceptable. A quick check of the hardware in use revealed that the DASD was outdated. Plans to upgrade the DASD had been postponed and the planned implementation was months away. We confirmed that the planned DASD would offer significant performance improvements, then worked with the responsible parties to expedite the upgrade. An overnight 30% reduction in batch processing time for the system resulted from this upgrade, with only a few hours of our time invested to determine the need and help bring it about.
Client: Mid-sized division of a large defense and commercial electronics company
Due to a change from fixed price billing for computer resources to "usage based" billing to be phased in over a 12-month period, the client was faced with a ten-fold increase in monthly charges for out-sourced processing of the Manufacturing Systems (primarily PIOS and Dispatch). Evaluation of current CPU usage revealed that the Dispatch reporting system, a vendor-supplied product, was consuming one third of the total time for Manufacturing. Consultants contacted the vendor and learned a new version was available that used significantly less system resources than the version currently in use. We contacted the appropriate support person, arranged to have the product upgraded, then made performance improvements to the few PIOS programs that were consuming excessive system resources. The net result was a reduction in billable resource usage for Manufacturing in excess of 20%. [Note: Actual monthly charges continued to increase due to the phase-in, but the monthly cost after the phase-in was complete was reduced by over 20%.]
A follow-on project involved targeting batch run-times, which were consistently exceeding the allotted batch window. Analysis revealed the primary culprit to be poor utilization of database buffers by IDMS batch programs. The release of IDMS in use did not support the flexible buffer sizes that were needed, but we had learned that the current release of IDMS had improved flexibility to optimize both batch and online buffers. We worked with the DBA to get the current release of IDMS installed, then to tune the buffers to more optimal sizes. The result was more than a 50% reduction in overall batch run-times, which not only eliminated the problem, but allowed significant room for future growth.
Copyright ©2002 Mainframe Software Innovations