performance hit in upgrade: 4.10 -> 7.20
Posted in 1999
Topics: Performance & Tuning, Installation, Setup & Upgrades, Connectivity: ESQL/C, 4GL & Embedded SQL, Platform-Specific Issues
Hello, My company has an application written in ESQL that ran in less than a minute on Informix 4.10. After upgrading to 7.20, the same code takes 2-3 hours. I am not too familiar with the code, as I started working here very recently. However, I do know (from sqexplain.out) that it is doing only selects, no updates or inserts. Also, many of the selects contain aggregate functions (max and count). Lastly, many (if not all) of the selects use prepared cursors that get prepared again and again as functions are called iteratively. Aside from rewriting the code, is there something that can be done to improve performance? I can't believe that the upgrade would cause such degradation, but I see it with my own eyes. There must be some factor that we have not considered... The O.S. is AIX 4.3. Thanks for any help. -David Dudas -----------== Posted via Deja News, The Discussion Network ==---------- http://www.dejanews.com/ Search, Read, Discuss, or Start Your Own
David Dudas wrote: > snip > > Aside from rewriting the code, is there something that can be done to improve > performance? I can't believe that the upgrade would cause such degradation, > but I see it with my own eyes. There must be some factor that we have not > considered... > Test with OPTCOMPIND = 0 (Environment or onconfig).
You should try running Update Statistics using the HIGH or MEDIUM keywords. It makes a lot of difference from just running Update Statistics by itself (which is defaulted to LOW). Check the performance guide for its recommendations for running Update Statistics. Take care. Clifton Markus Holzbauer wrote in message <36C3E583.708BA270@mch.siemens.de>... > > >David Dudas wrote: >> >snip >> >> Aside from rewriting the code, is there something that can be done to improve >> performance? I can't believe that the upgrade would cause such degradation, >> but I see it with my own eyes. There must be some factor that we have not >> considered... >> >Test with OPTCOMPIND = 0 (Environment or onconfig).