z/OS Performance Management: Humans Still Needed
EP Strategies' Scott Chapman highlights why a human eye is still essential when it comes to performance management in the age of AI
At Enterprise Performance Strategies, we help customers with their z/OS performance issues. While AI and machine learning have a place in augmenting mainframe technicians’ abilities, AI and ML are unlikely to replace humans in managing z/OS performance anytime soon.
The primary reason I say this is that (so far) we don’t have AIs that ask questions to understand your overall environment and your business objectives as they relate to your system and application performance. A simple example is capping: to lower their software costs, some customers intentionally limit performance to what is required to satisfy their business needs. Others avoid capping because it’s more important to them to provide as much capacity and performance as possible to their applications. Determining the proper capping level (if any) is a combined technical, financial and business decision.
We make many performance recommendations to customers, and these recommendations commonly come out of a larger discussion and almost always involve a trade-off of some sort. The most benign are those that redistribute resources from workloads that are providing more performance than is required to those that are suffering. But sometimes, harder choices must be made—some workloads might have to suffer for the greater good. Balancing competing workloads requires institutional knowledge.
In general, if you’re getting z/OS performance tuning recommendations, you should be able to understand both why the recommendation is being made and the side effects of the recommendation. AI and ML may be able to highlight areas for improvement, but it’s still up to you to understand your overall environment and what the real business priorities are.
Co-creator of Pivotor, Scott has more than 3 decades of diverse experience in mainframe performance and capacity, having been an application developer, application performance analyst, system programmer and capacity planner. At EPS, Scott has helped many customers understand and improve the performance of their z/OS systems.
See more by Scott Chapman
About Enterprise Performance Strategies, Inc.
Enterprise Performance Strategies, led by Peter Enrico, has assisted hundreds of installations with their z/OS performance needs. Learn more about our solutions →