Re: tune for INSERT performance?

Sanjay Mathew wrote:

>>

Try making the initial extent so large that the bulk inserts do not ask for space. This is the only thing that i can think of for improving the performance of the insert.
<<

Thanks for the advice. I neglected to mention that my 44 second test time was not the first test I ran. The first time I ran the test it took 70 seconds; subsequent tests took 44 seconds. This is because of the extents issue that you pointed out; on the first test the additional extents had to be allocated. Before I started the test the second time, however, I deleted all rows from the table. This freed up all the space for the second run to do INSERTs into - since the extents were already allocated, the test ran in 44 seconds instead of 70. I did try your suggestion, however, by dropping the table (which blows away the extents) and recreating it so that it had a small initial extent and then a 300KB second extent. The first test was about the same: 70 seconds; subsequent tests were around 44 seconds. So, I ended up with fewer extents (2 instead of 7), but the performance was the same, and is still slower than SQLBase.

>>

Other things too look for is the I/O contention on the disk where you are inserting.
<<

I shut down the other services on the system, and I was the only user connected to Oracle, so I think I've got this base covered.

Thanks,

Gregg Sporar
i2 Technologies Received on Mon Dec 22 1997 - 00:00:00 CST