Re: 8.1.7 cache buffer chains contention

On Tue, 28 Aug 2001 13:19:45 +0100, "Jonathan Lewis" <jonathan_at_jlcomp.demon.co.uk> wrote:

>
>The reported problem was 'cache buffers chains'
>latching, with one specific block being responsible
>for most of the lost time. This can only arise if
>(a) the latch is acquired too frequently for that
>buffer, or (b) the latch is held to long for that
>buffer.
>
>I doubt if LRU latching comes into it at all -
>although if the database write rate was too
>ferocious, then we would perhaps expect to
>see other waits appearing in association with
>this latch wait.
>

Yup, follow you. I was thinking more in line with a normal system where I'd expect a more even load. But this one is definitely abnormal. If the problem is the root index block, do you think a partitioned local index might help? I'm thinking: there would now be as many "root" blocks as there are local index partitions, with a consequent lightening up of the load on a single block. Does this make sense?

Cheers
Nuno Souto
nsouto_at_optushome.com.au.nospam Received on Tue Aug 28 2001 - 09:22:10 CDT