'Dear Rocky' answers your System z monitoring and performance tuning questions.
Rocky McMahan, a Senior Software Performance Engineer in z Systems Software R&D, offers valuable tips to help you get more out of your monitoring software.
I’m a Systems Programmer at a large financial institution – you may remember my question from several years ago regarding OMEGAMON for IMS. Our system data is very important to us, but we want a way to view and analyze this off-platform, utilizing enterprise standards such as Splunk, and open source solutions like Kibana.
I’ve heard about IBM’s Common Data Provider for z Systems, and would like to know how it utilizes my resources, such as my zIIPs. I’ve also see there has been a recent update in 3Q17 – what kind of improvements have been made in this update? Sincerely, -Awaiting in Amsterdam
Dear Awaiting in Amsterdam,
It’s good to hear from you again! I’m happy to answer your questions about the Common Data Provider for z Systems (or CDPz for short).
First of all, I think you’ll be very happy to know that CDPz can send data to both Splunk and Kibana! We believe that we should empower customers to use the solution that is right for them, which is why we’ve opened ourselves up to work with multiple data endpoints and there are more to come.
To answer your second question, CDPz will offload to your zIIPs! The Data Streamer portion of our product is a JVM, so it makes really good use of this feature. Just how much can be offloaded? Our tests have shown up to 80% of the Data Streamer process can run on zIIP, which can really save on CPU time!
Finally, we’ve made a few other changes with our latest release of Version 1.1.0 . First of all, we’ve set the default heap size for the Data Streamer JVM to 4GB. Working with our customers, we found that this was an appropriate default for most environments. Of course, this can be tweaked to fit your environment as needed. Next, we focused on other improvements to the Data Streamer to decrease the amount CPU seconds it uses. We’ve worked with customers and analyzed our code, ultimately making changes that reduce the CPU use of CDPz by up to 57%! Not only that, but we’ve identified changes in the SDE that handles SMF records that we’re working on for an upcoming PTF that should even further reduce CPU consumption! As always, we’re looking to answer more of your questions, and are working to deliver the best solutions to your problems!
Thanks to Chris Florence for contributing to this article.