Archives of the TeradataForum
Message Posted: Wed, 22 May 2002 @ 19:24:19 GMT
Experience unfortunately was the best teacher for us. We upgraded from V2R3 to V2R41 in March. I'd like to do several things differently the next go around.
Recommendations (sorry if they're long winded or redundant)...
- Keep several canned/standard reports as your benchmarks. Make sure you're comfortable with them (joins, tables, etc.). Pick reports that you've been tracking data rows returned/performance times for several months. In your tracking, don't forget to run EXPLAINs on these reports and keep the EXPLAINS on file with your rows returned/performance time history. These explains will come in handy when you least expect it.
-Talk to your power users and add some non-standard report queries into the benchmark mix. These queries are the ones that your users are running on a regular basis and would be familiar with the rows returned and performance. Don't forget to keep EXPLAINs on these specific user queries, too.
-Don't forget to run your reserved words checks at a minimum. Checktable and scandisk checks are good to run if it's feasible to do it on your system.
-It's good to run a test process on a separate development server first before implementing into production. When you run your test process, DON'T FORGET to 1)make a tape/system/database backup of your benchmark queries/explains and 2)keep an electronic copy/hard copy of the query results/explains.
-Include a data statistics strategy in your testing plan because V2R41 is stats hungry. Test out your data stats theories (refresh vs drop/recollect) in development first and look at the explain plans you generate. We unfortunately did not do this. Once the engineer finished the upgrade, he suggested that I refresh the statistics. I did the stats refresh and the explains/performance didn't seem to change much. Then I completely dropped and recollected statistics on the database. I couldn't say conclusively that the drop/recollect improved the performance either. Make sure you have the data stats strategy nailed down in development, first, before you do anything in production.
|Copyright 2016 - All Rights Reserved|
|Last Modified: 27 Dec 2016|