RE: 9.40 vs 9.21 performance

We had an issue which was may have only appeared in Solaris but we saw really slow performance over all. The problem was fixed in 9.30.FC7 (we have been confirming this in Dev/Test and plan to deploy ASAP but in the mean time the following helped with the issue. It was because the system was doing small updates, inserts, and selects. This was also suppose to be corrected in 9.4 since the code that causes this problem was migrated with no changes from 9.30 to 9.40. #### Set from ENV before Startup #### #### HELPED #### DONTDRAINPOOLS=1; export DONTDRAINPOOLS SPINLOCK_BREAKCNT=75000;export SPINLOCK_BREAKCN #### HELPED #### 9.3 and up (I cannot speak to 9.0, 9.1 and 9.2) all try to drain pools and maintain a cleaner set of resources for the database. Setting DONTDRAINPOOLS sets it back to the version 7 standard. Setting SPINLOCK_BREAKCNT allows the process to yield more often so other processes have more time to work. If there is a really high number of SPINLOCKS this could help. Defects Fixed in 9.30.XC7 release: ================================= 1. Bug Number: 162704 Short Description: PERFORMANCE DEGRADATION DUE TO MEMORY FRAGMENTATION AND ALLOCATION ISSUES Maybe checking the newly released software for this bug fix could help others. We almost rolled back and started with a migration project to another database engine because this was so bad and was not easy to reproduce. I am glad for development patches. Eric B. Rowell sending to informix-list