February 3, 2004 at 10:06 am
I am trying to scale an application by partitioning some large tables into smaller tables. I have a view which has a union all of 16 tables partitioned by clientID. When I do 1500 rows of updates using the view it takes more than 4 minutes to update. When I update using the base table it takes less than 30 seconds. The plan for the view shows all tables except the one being updated with 0 scan counts. The base tables have indexes. The update statement is first using the cluster and then the index based upon the clinet and id I am using to update the row. The plan for the table has a top 1 after the seek on the cluster. The plan for the view has a bookmark lookup. Should updates against the table and the view be about the same if using the correct indexes? What other items can I look at to partition this data?
thanks
February 5, 2004 at 11:13 pm
Add check constraint on smaller tables will help optimizer to search.
February 6, 2004 at 6:00 am
Check constraint is present.
Anything else? Is anyone using Partition Views to do updates or just selects?
Viewing 3 posts - 1 through 2 (of 2 total)
You must be logged in to reply to this topic. Login to reply