|
|
Archives of the TeradataForum
Message Posted: Wed, 16 Nov 2005 @ 12:09:45 GMT
Subj: | | Re: Excessive Delays between DBQLogTbl's StartTime and FirstStepTime |
|
From: | | Ballinger, Carrie |
Clay,
On this point of yours:
| What is strange, in our case, while a particular query is waiting to Start (i.e., Time > 'StartTime' and Time < 'FirstStepTime'),
many other all AMP queries will have already start and finished. | |
One thing that's easy to check is if the plans were cached for the queries that spent less time in the PE, and not for the query that spent
more time. Also, how complex was the plan was for the query with more PE time, if not cached, including were there a lot of join index choices
to be made. Parsing can take longer if there are lots of objects that need to read into the dictionary cache for a given query, compared to
queries have their dictionary information already cached. It's difficult to assess some of these things.
Probably these contrasts in PE time have always been there, it's just that we have not always had DBQL to observe them before.
Thanks, -Carrie
| |