Re: Update statistics for large databases

I believe you have a problem in that you set PDQPRIORITY to 1 before you do an update statistics on the procedures. Informix will run the procedures with the PDQPRIORITY that their statistics were updated with. So all of your procedures will run with PDQPRIORITY 1. Here is a quote from John Miller III: Attention: When using PDQPRIORITY with UPDATE STATISTICS, you do not want to accidentally compile your stored procedures with the PDQPRIORITY turned on. If you run UPDATE STATISTICS with PDQPRORITY set to 50 to build data distributions and compile store procedures, the stored procedures execute with the compiled time PDQPRIORITY, not the PDQPRIORITY of the user who executes the procedure. This is why it is recommend to always use the phrase "for table" when building distributions. The following statement will build distributions for the entire database: UPDATE STATISTICS high for table. http://www-128.ibm.com/developerworks/db2/zones/informix/library/techarticle/miller/0203miller.html hatricom wrote: > Hi, > > how can I obtain a short time for my update statistics on a large > databases. > It's run on a machine with 8 processors, 32 Gb Ram, RHEL4 and IDS 10 > FC4. > > My script for update statistics begins with update statistics medium on > database and continue with update statistics low and high for the > tables which are most used. > > I use PDQPRIORITY 80 for tables and 1 for procedures. > DBUPSPACE=25000:1000 > The script finish in about 5 hours, the database is 70 Gb large.