Re: Slower db file parallel read waits after migrating to SAN
- From: Adam Musch <ahmusch@xxxxxxxxx>
- To: nkodner@xxxxxxxxx
- Date: Tue, 23 Feb 2010 09:00:13 -0600
When you refer to a change in storage, are you referring to the
underlying storage, or did you move the tables and/or indexes
yourself?
If you changed the underlying storage, how so?
If you moved/rebuilt indexes, how did the statistics (specifically the
clustering factor of the indexes) change? Did you change anything
else about the optimizer statistics, such as which columns histograms
are calculated on and how?
Some other things to consider:
1. Partitioning. 422 million rows is a pretty sizeable chunk of
table, and I suspect list partitioning may be useful if you have fewer
clients with more skew in their volumes. If you have large numbers of
clients with fewer records per clients, hash partitioning may be
useful.
2. If you're going to redefine the index, I suspect the index should
be (CLIENT_NUMBER, ABAWD_STATUS_CODE, BENEFIT_MONTH). Depending on
the cardinalities of those columns (if the number of distinct values
for is "small") consider index compression, either for 1, 2, or all 3
columns.
--
Adam Musch
ahmusch@xxxxxxxxx
--
//www.freelists.org/webpage/oracle-l
Other related posts:
