Re: RA, bugs, and performance

Arun Shastry wrote: > > Carlson@WHSmith wrote: > > > Informix DS 7.23.uc4 > > HPUX 10.01 > > HP K200 4-way > > > > Up to this time, I have maintained a RA hit percentage of 90-99%. My RA > > variables were set to : > > > > RA_PAGES 150 > > RA_THRESHOLD 75> > > > While these numbers were a bit high, we were able to take advantage of > > most of the pages that were read-ahead. > > > > Wow, that's way too high for even a DSS environment. Rule of thumb I use > isto set these to about 32, 16 for a good sized DSS and to about 12,8 for an > OLTP envrionment. > Probably so . . . that was going to be part of my ongoing tuning, when the smoke finally clears. Even so, out RA% was about 95%. > > As part of my tuning, I was planning to revisit these variables. Where > > would be a good trade-off between read-aheads and bufwaits? I've seen > > the postings by various people, and took them to heart for the next > > round of R&D testing on my test system. > > > > Unfortunately, bug# 85350 changed all of that. > > RA parms. don't affect bufwaits, they are caused by not having sufficient > cleanbuffers available for a session. Check your cleaner activity, number > of buffers, > IO contention, LRU_MAX and LRU_MIN settings, and also your checkpoint > duration to get an idea of what could be leading to sluggish buffer > flusing. You > might also want to bump up your AIO VP's, if they are taking a beat. > IO could be an issue . . . I need to separate some hot tables. LRU_MAX and MIN might be an issue, currently set to 8 and 4. Number of LRUs could be an issue also . . . 3 CPUVPs, 8 LRUs, and 8 cleaners. > > > > > > I now have reset these parameters to 0. While I didn't mind retuning > > these variables, I hadn't planned on resetting them quite like this. > > > > Any idea of what the impact to my system would be? We're a mixed > > environment, OLTP and DSS batch reporting. > > I believe RA values are set a default, although I've never set them to 0 > :-) In order to get around bug 57338, not the bug number stated above, RA_PAGES and RA_THRESHOLD are both set to 0, thus disabling read-ahead. John Carlson Informix DBA WH Smith, Inc.