I completed my tests today. Same tests, the only difference being I used
two volumes that were not striped. With both index and table on the same
volume to insert 500k rows it consistently took twice as much time as when
the table and index were physically separated. I stand by the conventional
wisdom that says separating tables from their indexes yields a measurable
performance increase.
My average time to insert with both on the same disk was 1:14. With the
segments separated it was 2:30.
Received on Tue Apr 15 2003 - 15:31:56 CDT
