Dear all,
Kindly note that I had sometimes on AS/400, some jobs running in Qbatch and take almost the whole CPU for long time. Any other interactive job that should have few seconds, takes longer time to be completed.
Can we limit the CPU let say to 50% for Qbatch jobs?
Thanks in advance,
Wissam
The topic can be complex. Systems usually come from IBM configured for generic use. They work okay in most small-business environments with a wide variety of work types.
But they're essentially never configured to give good information about how the existing work management configuration should be changed to customize it for a specific small-business environment. Therefore, the initial technical steps involve changing a bunch of work management attributes so that appropriate details are reported to you. Those steps will not make things better, but they will get the system to start showing the kinds of changes that need to be made.
In order to do that properly (and to begin understanding the point behind each change), you need to learn Work Management. Your starting point will be the Work Management topic in the Information Center. Study the 'Introduction to work management', 'Concepts' and 'Managing work' sub-topics before starting anything. Make sure you know what those sections are saying.
If you haven't understood those sub-topics first, then it will be very difficult to describe the kinds of things you can do. It'll also be risky because some changes can make things worse, and you need to know what can be done about it.
A problem with auto-tuning is that the system's work management still should be properly configured first.
Essentially, each subsystem needs to have at least one additional memory pool assigned and its prestart and autostart entries need to be adjusted to point to the new pool. (There are a lot of entries.)
The chosen pools should be assigned appropriately. E.g., TCP/IP and its servers might be routed to *SHRPOOL1 while host server jobs might use *SHRPOOL2. Application batch might be *SHRPOOL3. Those are examples for discussion only.
Initial configurations have almost everything running in *BASE, except for QINTER and QSPL jobs. But having active work running in *BASE effectively negates automated tuning efforts. And having so many jobs running in the same memory pool makes it practically impossible to determine which ones of all the jobs are causing excessive faulting or other issues.
When a single pool (*BASE) has most jobs running in it, there is very little practical tuning that can even be done. Running an autotuner might do little but add unnecessary CPU cycles.
That's the kind of reason that an understanding of work management is needed. A good product will guide through the steps, but it can't make good decisions for you about the work on your system. An initial configuration should be set whether a product is used or not.
Free Guide: Managing storage for virtual environments
Complete a brief survey to get a complimentary 70-page whitepaper featuring the best methods and solutions for your virtual environment, as well as hypervisor-specific management advice from TechTarget experts. Don’t miss out on this exclusive content!
Discuss This Question: 7  Replies
The quick answer is yes, you can do that. There are many ways as to how to do it.
I recommend you study performance tuning on theAS400. You need to look at timeslice, pools, priorities, and other functions.
Thanks Charlie...
do u have a pdf or website link for this...
Regards,
Wissam
Dear Tom,
i have one M50 and one M25.
thanks
The topic can be complex. Systems usually come from IBM configured for generic use. They work okay in most small-business environments with a wide variety of work types.
But they're essentially never configured to give good information about how the existing work management configuration should be changed to customize it for a specific small-business environment. Therefore, the initial technical steps involve changing a bunch of work management attributes so that appropriate details are reported to you. Those steps will not make things better, but they will get the system to start showing the kinds of changes that need to be made.
In order to do that properly (and to begin understanding the point behind each change), you need to learn Work Management. Your starting point will be the Work Management topic in the Information Center. Study the 'Introduction to work management', 'Concepts' and 'Managing work' sub-topics before starting anything. Make sure you know what those sections are saying.
If you haven't understood those sub-topics first, then it will be very difficult to describe the kinds of things you can do. It'll also be risky because some changes can make things worse, and you need to know what can be done about it.
Tom
Here is another link for you.
https://search400.techtarget.com/feature/Fast-guide-to-improving-iSeries-performance
You may also want to consider buying some software that will monitor your system and do auto tuning for you.
...and do auto tuning for you.
A problem with auto-tuning is that the system's work management still should be properly configured first.
Essentially, each subsystem needs to have at least one additional memory pool assigned and its prestart and autostart entries need to be adjusted to point to the new pool. (There are a lot of entries.)
The chosen pools should be assigned appropriately. E.g., TCP/IP and its servers might be routed to *SHRPOOL1 while host server jobs might use *SHRPOOL2. Application batch might be *SHRPOOL3. Those are examples for discussion only.
Initial configurations have almost everything running in *BASE, except for QINTER and QSPL jobs. But having active work running in *BASE effectively negates automated tuning efforts. And having so many jobs running in the same memory pool makes it practically impossible to determine which ones of all the jobs are causing excessive faulting or other issues.
When a single pool (*BASE) has most jobs running in it, there is very little practical tuning that can even be done. Running an autotuner might do little but add unnecessary CPU cycles.
That's the kind of reason that an understanding of work management is needed. A good product will guide through the steps, but it can't make good decisions for you about the work on your system. An initial configuration should be set whether a product is used or not.
Tom