During search400's online event "iSeries performance: Time for a tune-up" with John Brandt from iStudio400, we received several questions from attendees that we didn't have time to address during the broadcast. John was able to answer some of those questions offline, and we've posted them here.
If you have more questions for John, you may send them to us at email@example.com.
Q: I just received this message on my AS400 (720) running V4R5M0: Interactive activity approaching capacity of installed. Are there any adjustments that can be made in response to this message?
A: This message is based on your OS/400 license. The system has determined that you have almost reached the total number of interactive users that your system is licensed to serve. Increase your licenses and this message will go away. There may be a threshold value somewhere in the system that controls when the message is issued. I have an open license, so I can't tell you where that value might be.
Correction: First of all, thank you to Chris Kennedy for pointing out that the answer I gave to a question was incorrect. I would also like to say thank you to Mr. Richardson for providing me with some detailed information on resolving this question and the problems these jobs have presented.
The real answer should be that the CFINT and CFINTnn jobs that run are governor jobs that restrict the CPU usage. These jobs have been on the system for a few releases, but have been tightened up by IBM as a result of some software that allows users to fool the system licensing software. The CFINT jobs will slow down the system based on the OS/400 CPU usage license. Chris pointed out, "... on a 730 (2067) processor with an interactive card of 1510, your interactive capacity is 1050CPW/2000CPW X 100 = 50%. This means that your interactive jobs are allowed to use 50% of your CPU's capacity for interactive requests. Whether it's one user or 500 users does not matter!! As you add more users to your system, it stand to reason that interactive use of the CPU will rise, but there is no limit on users."
I would also like add that part of the "tightening" of these jobs is to prevent software from "fooling" the licensing code. Mr. Richardson was very helpful in answering this question. During our conversation, he advised me that he has been in a tuning mode for the last few months and has found that the following steps will assist in controlling these jobs in a medium to large system environment (over 300 interactive users).
1. Change the interactive jobs to prevent purging to the *BASE pool. - PURGE(*NO)
2. Separate interactive pools to allocate specific memory pools to groups of users.
3. Separate batch pools where practical to speed those processes.
4. Use the ASP spreading values to allow the system to spread disk usage across the system.
He reported that these steps have significantly reduced the response times for users, increased the speed of batch processes and reduced the amount of CPU used by the CFINTnn jobs in his system. I stand (sit) corrected and encourage any user to correct me at any time.
Q: I have an ODBC request that performs the "Select" stmt. The message I get is that the query optimizer has timed out.
A: Go to the ODBC setup and increase the record blocking to the largest value possible. If that doesn't help, let me know and we can see what other things can be done.
Q: There is a performance monitor collection base on the AS/400 and a part of Service Director. What is it collecting and how can we make use of it?
A: The performance monitor processes allow you to track what jobs are taking the most CPU and resources on your system. The guide is in the AS/400 online library. It can detail what you get and how best to use it. Beware that turning it on will slow your system significantly while it is in the collection mode.