Re: Defrag 8i DB

"Paul Drake" <drak0nian_at_yahoo.com> wrote in message news:1ac7c7b3.0402201325.6f89e393_at_posting.google.com...
> Howard,
>
> Oracle using an extent size of 64KB as the starter in auto-allocate
> has me wondering. I chose 128KB extents (uniform) way back when we
> migrated from 7.3.4.5.x to 8.1.6.3.0. I shudder to think that I should
> perform a re-org, just to reclaim 32 MB of space, for those segments
> that didn't allocate more than 8 (8KB block size) blocks.
>
> Our default db_file_multiblock_read_count was usually 16, so 16 * 8192
> = 128KB.
> It certainly kept the math easy.
>
> In Dell-land, their PERC raid controllers had a max stripe size of 256
> KB.
> In our environment, setting dbfmbrc = 32 caused the CBO to think that
> full table scans were too cheap, so 128KB reads for fts seemed okay.
> It certainly deserved more testing than I gave it.

Same here!

> so now that we're getting around to testing 10.1.0.2 (on RHEL 3 ES), I
> guess I'll look at using auto-allocate with automated segment space
> management, for everything that is not clustered. I don't see me using
> anything other than uniform-sized extents with clusters, but I also
> haven't read anything regarding storage parameter impacts on index and
> hash clusters in 10g.
>
> Pd

Well, you know my opinion of ASSM in 9i. I have yet to check whether it's gotten any better in 10g.

I wish they hadn't called 10g's ASM (which seems like a good idea) practically the same name as 9i's ASSM, which is a pig.

:-)
HJR Received on Fri Feb 20 2004 - 15:48:47 CST