Tweet
We upgraded from v4r2 (!!!) to v5r3 during the first week of July of this year. I had heard some rumors that performance was an issue with some folks after the upgrade, but had a 2nd test system that it was running on (of exact same configuration) & it did seem to take a little longer with our nightly jobs, but nothing dramatic. We have about 100 batch jobs that run each night & although a few of them seem to run a lot longer, 25 minutes as opposed to 15 minutes, there as of yet hasn't been anything that really negatively impacts overall processing. Online response time, which I was mainly worried about, didn't seem to be affected, at least not enough that I noticed or any of my users have felt compelled to complain about so far, and they are not shy about complaining 
. Our hardware is relatively old too, a 170 with 210 CPW batch & 29 CPW interactive. We had 5 8.5 gig 6713 disks in the system at the time of the upgrade, all of them in a parity set (IBM's term for RAID-5, I believe). The upgrade itself took up 16% of the available DASD coming from v4r2 so 2 weeks ago I installed 5 17.5 gig 6718 drives & added them to the system ASP in their own parity set. Performance-wise, these extra disks didn't have any noticable ill effects either. Total usable disk space is now 104 gig. One thing I did notice was that the IPL took a LOT longer to complete. Same amount of time for the initial sign-on screen to appear on the console, about 15 minutes, but the system was just THRASHING for another 45 minutes with response time being pretty bad. I noticed that there was a job taking up a lot of the CPU % during this time & IBM advised me that it was for clustering. Since we don't have a cluster of AS400's, I removed this job's prestart or autostart job entry from the QSYSWRK subsystem description & now IPLs finish like they used to pretty much. It made a big difference anyway. I forget the job's name but I remember it had CST embedded in it. If your not using clustering, maybe you could call IBM & get them to help you remove this job from the autostart list. HTH, Martin
. Our hardware is relatively old too, a 170 with 210 CPW batch & 29 CPW interactive. We had 5 8.5 gig 6713 disks in the system at the time of the upgrade, all of them in a parity set (IBM's term for RAID-5, I believe). The upgrade itself took up 16% of the available DASD coming from v4r2 so 2 weeks ago I installed 5 17.5 gig 6718 drives & added them to the system ASP in their own parity set. Performance-wise, these extra disks didn't have any noticable ill effects either. Total usable disk space is now 104 gig. One thing I did notice was that the IPL took a LOT longer to complete. Same amount of time for the initial sign-on screen to appear on the console, about 15 minutes, but the system was just THRASHING for another 45 minutes with response time being pretty bad. I noticed that there was a job taking up a lot of the CPU % during this time & IBM advised me that it was for clustering. Since we don't have a cluster of AS400's, I removed this job's prestart or autostart job entry from the QSYSWRK subsystem description & now IPLs finish like they used to pretty much. It made a big difference anyway. I forget the job's name but I remember it had CST embedded in it. If your not using clustering, maybe you could call IBM & get them to help you remove this job from the autostart list. HTH, Martin
Comment