Help with TASM and Tactical queries

Database

Help with TASM and Tactical queries

We have a mixed workload that includes web queries (customers), ETL jobs, and BI reporting users.  When there are no ETL jobs or BI reports running, our web (tactical) queries average 2 seconds.  When other workloads are running, especially long-running jobs, our tactical performance is terrible (30 - 60 seconds).  I've used TASM to give 90% priority to tactical and I've limited the other workloads to 50% CPU. Does anyone have any suggestions for protecting the tactical queries from the other workloads?

2 REPLIES
N/A

Re: Help with TASM and Tactical queries

There might be many other options but I guess the forum will have limited chances to help here. At least there would be some danger for your system in respect to SLAs if you just follow some quick guesses based nearly no input from your side.

Setting up TSAM correctly requires a detailed analysis of the current workload and the SLAs to fulfil. It usually requires some days to gather and analyse the info and come up with a good recommendation. I know TD offers support there and you might find also some externals which will help you.

Beside this - now we come to the quick answers based on guessing ;-> - 2 sec avg for tactical sounds long to me which is indication all amps queries. This can caus amp worker tasks shortages etc. So you might have some room for improvement by analysing and improve performance of the tactical queries.
Teradata Employee

Re: Help with TASM and Tactical queries

It is tough to guess as to what exactly is going on but some thoughts. Is there possibly a lock situation here where the tactical queries are getting blocked? Are the amount of long running jobs a big number? If so, you may want to consider throttles for the longer running jobs. Also what is the overall WD strategy you have setup as you may need some finer granularity. Also have you tried a smaller relative percentage for the longer running queries? Like 90% tactical, 10% to long running jobs? Are the longer running jobs experiencing nasty skewing? If so, you may want a TASM exception to track this and either fix the SQL or have those jobs downgraded to a lesser percentaged WD. There really is no one size fits all solution here as the solution is very case dependent with a number of variables to consider.