Re: Redo latch tuning question

I'll bet the absolute number of gets/misses is tiny compared to the immediate gets/misses.

On the redo copy latch the 'willing to wait' gets/misses can be ignored.

--

Jonathan Lewis
Yet another Oracle-related web site: http://www.jlcomp.demon.co.uk

Anne Hanson wrote in message <37F4C91B.E1FEC77A_at_fas.harvard.edu>...
>On one of my servers, I have heavy contention for redo copy latches. The
>gets/misses ratio is extremely poor. The immediate gets/misses ratio is
>fine. The redo allocation latch is not experiencing any contention. I'm
>running 8.0.5 on NT 4.0. I've been reading up on latches, and the two
>solutions I've seen suggested are:
>
>1) Increase the LOG_SIMULTANEOUS_COPIES initialization parameter to 2 *
>number of CPUs. Right now it is 2, as the server has 2 CPUs.
>
>2) If the above doesn't help, increase the LOG_SMALL_ENTRY_MAX_SIZE so
>more redo entries will be handled by the redo allocation latch.
>
>Does anyone know of any down side to either of these solutions, or
>anything else I should be thinking of? Thanks in advance for any tips,
>
>Anne
>
>
Received on Fri Oct 01 1999 - 12:21:45 CDT