Mainframe, MVS and zOS Discussion

Hi,

I have some idea about the capacity planning and management in linux and unix. But now I have been asked to look for the mainframes also. One concept which I tried to explore is: memory leaks. Does memory leaks happen in mainframe too? And if they do, how do you recognize them? If you were given the log of 100 of mainframe serves with historical data, with CPU and memory usage is there a way to know which server can have a memory leak problem?

Looking forward to your valuable replies.

Yes, memory leaks happen to mainframes. However, you CANNOT take a snapshot at a point in time and state categorically that this mainframe has a memory leak. Memory leaks are found by checking memory usage over time; if the common memory usage keeps increasing over time then there may be a memory leak, subject to confirmation. Mainframes also have different pools of memory (CSA, ECSA, SQA, ESQA for example) and a memory leak may only occur in one of them. If your site has a monitoring tool such as OMEGAMON or MAINVIEW, these are quite helpful in identifying memory leaks. A typical mainframe is running 100 to 200 address spaces (started tasks, TSO users, batch jobs) at any given time (and big mainframes can be running hundreds more), so identifying the source of a memory leak can require a lot of investigation.

At a previous employer, we had a memory leak in a third-party software product. We knew which product but we couldn't get permission to install the latest release (which hopefully would have fixed the memory leak). However, since the leak was slow (a few K a week) and easily managed by an IPL every 3 months, we managed it that way.

Robert Sample wrote:Yes, memory leaks happen to mainframes. However, you CANNOT take a snapshot at a point in time and state categorically that this mainframe has a memory leak. Memory leaks are found by checking memory usage over time; if the common memory usage keeps increasing over time then there may be a memory leak, subject to confirmation. Mainframes also have different pools of memory (CSA, ECSA, SQA, ESQA for example) and a memory leak may only occur in one of them. If your site has a monitoring tool such as OMEGAMON or MAINVIEW, these are quite helpful in identifying memory leaks. A typical mainframe is running 100 to 200 address spaces (started tasks, TSO users, batch jobs) at any given time (and big mainframes can be running hundreds more), so identifying the source of a memory leak can require a lot of investigation.

At a previous employer, we had a memory leak in a third-party software product. We knew which product but we couldn't get permission to install the latest release (which hopefully would have fixed the memory leak). However, since the leak was slow (a few K a week) and easily managed by an IPL every 3 months, we managed it that way.

Thank you Robert! How did you identify that there was a leak of few Ks a week? What tool have you used or what made you believe it was a memory leak?

We had BMC MAINVIEW installed, which allows you to look at which programs (address spaces) have how much memory in the various pools. This allowed us to pinpoint, over time, which vendor program had the leak.