Re: Forall performance

Well, suppose you can save 5 seconds from reducing context switch by using forall, but you waste 5 minutes due to inefficient redo log management, you will not notice the benefits of forall. The larger the data volume the more likely you will suffer from inefficient redo log management. That was my guess as to what might be going on behind the scene. Just a guess.

But, the op's 3*10 meg redo logs seem big enough to me. I would take a look of the rollback segments as well. Or maybe the optimal array size is system dependent - as I seem to recall reading it somewhere at Metalink.

In any case, the direct insert hint is missleading and should be removed.

• Jusung Yang

dyou98_at_aol.com (D.Y.) wrote in message >

> I think the OP's question is why it's necessary to have an array size
> of 1000, 2000, etc. to get significant performance improvement from the
> FORALL statement. Well, the main reason bulk binding can speed things
> up is because it reduces context switching by binding multiple rows at
> the same time. A one row array really defeats this purpose. Incidentally,
> I too have found 1000 to be close to the optimal size.
>
>
Received on Fri Feb 14 2003 - 21:40:06 CST