IOT or Cluster or neither
- From: Aditya Alurkar <aalurkar@xxxxxxxxxxxx>
- To: <Oracle-L@xxxxxxxxxxxxx>
- Date: Fri, 18 Aug 2006 07:12:20 -0700
I am evaluating a good table design to store data with the following
characteristics.
* Large number of records >200million rows
* Tables have a unique "id" for each row (generated using sequence).
* The rows can be logically grouped per "user_id"
* The rows are primarily accessed per "user_id"
* The distribution of the rows is not even among "user_id" i.e. Most users
have <50 rows but there are a few who have >10000 rows.
* The tables need to have some secondary indexes.
* The tables have "minimal" updates and "moderate" inserts.
In order to increase the performance I have the following options.
A) IOT with primary key similar to "userid,id"
B) Cluster the table on user_id (index)
C) Cluster the table on user_id(hash)
Since the distribution of the data over "user_id" is not equal I'd like to
get feedback from folks who have found themselves in similar situations and
what solutions they have chosen and why.
Appreciate any input/comments
--
Adi Alurkar
aalurkar@xxxxxxxxxxxx
--
//www.freelists.org/webpage/oracle-l
Other related posts:
