Re: How to "Defragment" Tablespaces ?

Nuno Souto wrote:
>
> On Mon, 9 Apr 2001 09:48:29 -0400, Liz Reen <lizr_at_geologist.com>
> wrote:
>
> >
> >A tablespace is much like a disk drive. Think of tables as files. Like
> >files tables have extents all through the tablespace. There has to be a
> >diskmap sort of data structure which keeps track of the extents and
> >tables. This is where the contention happens. I had heard that there
> >was a limit of approximately 25 concurrent accesses allowed to this area
> >in 7.x. I do not know about 8.x.
>
> Minor correction. With DB2 and other databases, there is a tendency
> to consider a tablespace as a "file". That is not the case with
> Oracle. You can have a heap of files making up a tablespace.
>
> But your point about "there has to be a diskmap ..." is quite correct
> and indeed there can be contention for the equivalent in Oracle. And
> even with EMCs, it will make a difference. In fact, it is my
> experience that it is *precisely* with EMCs that you notice this the
> most, but unfortunately I haven't yet had a chance of proving it
> beyond any doubt.
>
> Cheers
> Nuno Souto
> nsouto_at_bigpond.net.au.nospam
> http://www.users.bigpond.net.au/the_Den/index.html

As further reinforcement you can see a 'bug' on Metalink where a large Oracle user (who I happened to be working for at the time) got massively hit by "performance problems" when upgrading to 8i - and NONE of the explain plans had changed...the reason was due to a change in the calls made to the underlying disks (in a veritas file system) which brought the database to a standstill on internal locking issues (a variant on inode locks)...the only fix was to not use vxfs file systems (by moving to quick io)...

Another reason I prefer to use raw on locally managed tspaces..

Cheers
Connor

-- 
===========================================
Connor McDonald
http://www.oracledba.co.uk (mirrored at
http://www.oradba.freeserve.co.uk)

"Some days you're the pigeon, some days you're the statue"