Archives of the TeradataForum

Message Posted: Fri, 13 Jan 2012 @ 21:17:09 GMT

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

Hi all,

i'm currently thinking about a modified way (in TD13) to determine when to recollect stats on large tables based on row counts and i would like to ask for your comments.

#1 An SP with a cursor based on my stats query with some filtering criteria sorted by tablename.

For each table (small/large based on the row count from old stats)

- small table: recollect stats on table level large table:

- collect PARTITION stats

- extract the new row/partition count with a single AMP PI-access

- row count (per partition) changed significantly?

- recollect, else skip the table

- for partitioned tables: number of partitions changed?

- recollect, else skip the table

Did anyone try something similar already?

#2 Did anyone test running multiple column/index level COLLECT STATS on a large table in parallel vs. recollection on table level?

There's a kind of batch window for recollecting, so resource usage is not an issue and submitting the collects in parallel is easy using a queue table.

Dieter

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