ray,
> Are you taking full advantage of you physical memory? How does you
> caching ratios look (onstat -p)? If not in the 90's play arournd with
caching ratios are good, high 90's.
> BUFFERS and SHMVIRTSIZE. Also, is the engine frequently allocating
> additional shared memory segements? If so, this will impact performance.
> Once again SHMVIRTSIZE may be low.
i'm allocating at startup, and during the day, so i bumped SHMVIRTSIZE
way up. i will see how it does when i reboot.
>
> If your checkpoints are frequently taking 4 seconds to complete, I
> believe that is too long. Play around with LRU parameters. Maybe you
most of this had to do with databases having no stats. i had too
many buffers for my usage here as well. between doing the update statistics,
lowering my number of buffers, and dropping LRU_MAX/MIN_DIRTY, my checkpoints
are now in the 0 to 1 second range, at least 80% of the time.
thanks for the response.
mickm