Archives of the TeradataForum
Message Posted: Mon, 23 Jun 2003 @ 16:13:22 GMT
| Subj: | | Re: Statistics didn't change |
| |
| From: | | Geoffrey Rommel |
Maeva, we may need a little more information.
| | I work with federated systems. | |
Are you referring to 'federated' as defined in IBM DB2/UDB?
| | in each data sources (DB2, | |
Is this DB2 for z/OS (mainframe) or for Unix (AIX, Linux, etc.)?
| | After, I executed "collect statistics on column" then "collect statistics on table" . I re populated my nickname. I compare the
new explain plans, but for Teradata, nothing has changed!!! Teradata after gathering statistics still thinks I have 1000 rows (I only
inserted 100 rows). | |
Well, in Teradata, the difference between 100 rows and 1000 rows is trivial; it would probably not be enough to change the query plan.
Still, if the stats were collected correctly, they should be correct. (Dite-la trois fois rapidement!)
'Nickname' is not a Teradata term, but it is used in DB2/UDB. Are you running these statements through an IBM tool of some sort? It may
have used improper syntax.
| | But do I need to collect statistics on it [the index] when it is created by default? | |
Yes. Statistics are null until you run "collect statistics".
By the way, this statement:
collect statistics on [database-name].[table-name];
re-collects only stats that have already been collected. If there were no stats before, there will be none afterwards.
| 
