Archives of the TeradataForum

Message Posted: Mon, 13 Nov 2000 @ 16:28:31 GMT

	<Prev	Next>		<<First	<Prev	Next>	Last>>

A couple fo things to consider:

First, it is important to always keep the weightings assocaited with the performance groups is the descending order from high to low, as it is initially set up in the DEFAULT partition. For example,

     PG#     Weighting
     7       40              Rush
     6       40              Rush
     5       20              High
     4       20              High
     3       10              Medium
     2       10              Medium
     1       5               Low
     0       5               Low

I believe Carrie addresses this issue in the white paper. The Priority Scheduler relies on the weighting being in this order for several reasons.

Second, I don't know what version of software you are on, but there is a know problem in using Priority Scheduler having to do with certain types of workloads, particularly very CPU intensive activity like product joins. This type of workload can starve other types of work from getting any CPU time. DR 52503 discusses this problem and the root cause. There is a fix for it in the following release:

     PDE     03.00.03.31
     DBMS    03.00.03.63

It's hard to tell if this is, in fact, the problem you are seeing. But if the big insert selects or joins are very CPU intensive and the other work is in a performance group with an allocation group policy of DEFAULT, this patch could address your problem.

Regards,

Tom Stanek
TFS Consulting

	<Prev	Next>		<<First	<Prev	Next>	Last>>