Ways to check the performance of a program?

[Amey K]

Hi,

I've coded some programs. Now what ways do you suggest to check the performance of a program? And if some programs are rewritten to replace some old code? Please share your views.

[Robert Sample]

There are tools such as STROBE that can be used to check the performance of programs, or you can use a monitor such as MAINVIEW or OMEGAMON to look at the program performance while it executes. If your site does not have such a tool available, then you probably won't be able to do much analysis as such tools typically cost hundreds of thousands of US dollars.

However, there is critical question that needs to be answered -- WHY are you concerned about program performance? These days a z/OS system can execute millions to hundreds of millions of COBOL statements per second of CPU time (depending upon the hardware used). Unless you have a known performance issue (such as your batch jobs are not completing within your batch window), there is really no reason to be concerned about performance of a program.

And rewriting old code is not generally cost-effective since a lot of testing is required to validate that the business logic has been correctly transferred.

[Amey K]

Thanks Robert.

However, there is critical question that needs to be answered -- WHY are you concerned about program performance? These days a z/OS system can execute millions to hundreds of millions of COBOL statements per second of CPU time (depending upon the hardware used). Unless you have a known performance issue (such as your batch jobs are not completing within your batch window), there is really no reason to be concerned about performance of a program.

If it is so, why do they keep on talking about MIPS optimzation all the time???

[Robert Sample]

Some part of it is sites that are running out of CPU (for whatever reason). Some part of it is inertia -- CPU used to much more important relatively when the machine cost a lot compared to programmers. Some part of it is a desire to get the most benefit from the very visible high-dollar mainframe. And part of it is organizations that don't realize they are spending thousands of dollars of programmer time to optimize code that will save them cents or dollars overall (or even nothing if they purchased the machine).

Really these days, if a site has a performance issue (such as not being able to complete the batch work in the batch window, or -- as happened at one site I worked at -- the time to complete the daily workload ran 33 hours) then there needs to be a focus on optimizing code. However, if a site is completing their batch work in the batch window and have no known performance issues then spending programmer time (and organization money) on improving program performance is generally a waste of resources.