Archives of the TeradataForum

Message Posted: Thu, 26 Apr 2007 @ 09:33:25 GMT

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

Hi Nazy,

I'd be a little careful about statistics collection on dictionary tables, and do it on an as-needed basis only. Where it could be useful is if there are frequent DBA or end-user queries against the specific dictionary tables, particularly if frequent joins are involved and the table are large. Then, the same considerations and trade-offs would apply under those conditions as would apply against user table statistics collection decisions (performance improvement of the queries, size of the tables, frequency of execution of the queries, time involved to collect/refresh stats).

I'd keep away from collecting statistics on highly volatile dictionary tables, such as DBQL or other data collection tables. A better approach with those tables is to offload them frequently, and consider index and statistics needs against the offloaded version of the data.

One thing to remember, some of the dictionary tables are non-hashed, such as Acctg and DatabaseSpace. If you try to collect statistics on a column belonging to a non-hashed table, such as DatabaseSpace you will get this error:

3876: 'databasespace' must be a hashed table.

Thanks, - Carrie

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