It’s free. It’s simple to enable. It can improve the performance of your IBM i system.
Many of you are already taking advantage of Expert Cache, but I still occasionally encounter customers who aren’t using it and are unaware of the potential benefit it can provide. If you’re one of these people, please read on to learn about a simple, no-cost capability that can improve your system performance.
So what is Expert Cache? It is an IBM algorithm that applies intelligent methodology to dynamically determine which objects should have larger block sizes brought into memory, and how long objects remain in memory. What you really need to know is that enabling this functionality provides you with the capability to reduce the number of disk I/Os a job has to do. For a batch job, this means it may be able to complete work in less time. For interactive jobs, it can mean faster response times.
Enabling Expert Cache is easy. It’s an available option for all shared memory pools (except the machine pool. If you have private pools configured, these pools are not eligible to use Expert Cache. From the IBM Navigator for i interface, navigate to Work Management –> Active Memory Pools. Then right-click on a Shared Pool, select Properties, and click on the Configuration tab. The default for “Paging option” is Fixed. Change this to Calculated, then click OK. Expert Cache is now enabled for jobs in that pool:
If you’re a green screen user, this option can be changed on the Paging Option column on the Work with System Status (WRKSYSTS) or Work with Shared Pools (WRKSHRPOOL) screens.
If you’re wondering which of your jobs might benefit from Expert Cache, check out the Performance Data Investigator (PDI) capability in Navigator (Performance –> Investigate Data). Several available charts in Collection Services can provide insight as to which jobs are doing the most physical synchronous reads. An example is the Physical Disk I/O by Job or Task – Detailed chart:
With Expert Cache enabled, you could expect to see a reduction in the amount of physical read I/Os and a larger read I/O size. However, benefits from Expert Cache can only be realized when memory is not constrained and the CPU is not over-utilized.
If you’re not currently taking advantage of Expert Cache, give it a try. You may find yourself with a better performing system.
This week’s blog post is written by Stacy Benfield, a member of the Lab Services Performance team. Previously, Stacy was a member of the IBM i development team, spending time on both the Work Management team as well as leading the Performance Data Investigator team.
This blog post was originally published on IBMSystemsMag.com and is reproduced here by permission of IBM Systems Media.