Re: resolving buffer busy waits

interesting. will look into it. indexes have bbw's as well. but they are "minor" compared to the table waits. want to tune the prolific ones now (which are on the tables) and attack the indexes later. reverse keys will be helpful there, but it is not _my_ app, so must tread carefully.

thing is, rebuilding to a hash partition is a rebuild and requires more time. increasing the buffer cache/"keeping" the hot tables is easy. but perhaps i am being naive. IO ain't great and i fear the extra buffers will impact dbwr/checkpointing and, in turn, the app. so the problem just sort of shifts ...

another comment to anyone interested - wouldn't resolving 130 mode bbw's using a keep buffer pool, but not data density (rows per block), simply shift the bbws mode 120 or some other wait? if so i assume the wait time would diminish, so potentially helpful, but not the "final resolution"???

>
> hash partition the hot table? That should work in 8.1.7.3 IIRC.
> Don't forget the indexes as well.
>
Received on Mon Sep 15 2003 - 04:28:34 CDT