Re: switching to locally managed tablespaces

It is definitely possible that I could have said that, but it was a typing error if I did. From a clean install there is one bitmap per file.

With an 8K block size, the typical header will be 64K, of which 2 blocks are not bitmap, so you get 48K of bits, or 384,000 bits - so the limit is high, but not extreme.

So if you do have lots of objects which are all rapidly extending because they all have a (relatively) small you have at least three overheads to consider:

1. scanning the bitmap for a free bit
2. the contention from multiple processes wanting to update the same bitmap
3. the redo generated by the bitmap update

These are probably likely to be pretty insignificant.

Updates to TSQ$ if you have quotas imposed, will probably be a more important concern.

The fact that a segment header at 8K only maps 505 extents may have some performance side effect.

But to me, one of the most important issues is the administrative one. If you want an email sent to you whenever an object grows - just in case of a space threat - you need the right sized extent because the "right" email is the one that:

1. doesn't appear very often
2. has just a couple of lines when it appears
3. is full of nothing but important content.

If you get a report every day that say 75 objects have all extended by somewhere between 5 and 50 extents, you stop seeing what's in front of you.

--
Jonathan Lewis
http://www.jlcomp.demon.co.uk

Next Seminars
        UK            July / Sept
        Australia      July / August
        Malaysia        September
        USA (MI)        November
http://www.jlcomp.demon.co.uk/seminar.html

Niall Litchfield wrote in message
<3d22b879$0$230$ed9e5944_at_reading.news.pipex.net>...


>

>Jonathan pointed out earlier (if I read him correctly) that there is a 'per

>tablespace' limit for the number of extents which will fit in the bitmap.

So


>you may not get degradation on any individual table with a few thousand

>extents but if the bitmap is dealing with hundreds of thousands of extents

>it may suffer.

>