Re: Large Index organized tables: just a marketing feature ?! (for experts)

At present I am testing import and "insert into select from". I am getting closer to have a good load time .

I have figured out that the order of the data in the source (table or file) has an important effect on the load performance of the IOT table. I have an index splitting problem with IOT. On the B-tree index there is no index splitting. I don't know if it is a bug or a feature.

After three hard weeks of work finally I am getting a solution: (46 000 000 row sample)
1. I load the data into a regular partitionned heap table. 2. I create a b-tree index on the table. I am using the same keys as the primary key: product_id, timestamp, .... 3. I load the data using "insert into select from".

Everything is done within an hour (instead of 9 hours). Maybe this information can help somebody later .... Received on Thu Jul 31 2003 - 06:18:27 CDT