Re: DB_BLOCK_SIZE set to the highest value

Interesting;

There is a feature of rollback segment headers that is not often mentioned - the 'free block pool' that, loosely speaking, lists the last record position of the most recent five transactions to commit.

Having this list allows Oracle to step backwards (to a very limited degree) in the rollback segment and reuse space that would otherwise be wasted.

If you persistently run with more than 5 concurrent transactions per rollback segment, and the transactions are all very short, you can waste blocks very rapidly - and bigger blocks waste more space.

Might this have something to do with your problem ?

--
Jonathan Lewis
Yet another Oracle-related web site:  http://www.jlcomp.demon.co.uk

Practical Oracle 8i:  Building Efficient Databases
Publishers:  Addison-Wesley

Reviews at: http://www.jlcomp.demon.co.uk/book_rev.html



Nuno Souto wrote in message <3b0e6eef.19799819_at_news-server>...


>

>I'm sure it was because of the particular characteristics of the

>application in question rather than this being wrong.  Bear in mind

>his points are somewhat general.  That means: they may not apply to

>EVERY situation.  In my case, a 16K block size *consistently*

>generated snapshot too old messages. To the point where half the batch

>programs were being affected by it.  And I needed nearly 8 Gb of RLB

>size for a total database table size of 15Gb and the largest table at

>300Mb!

>

>Changing to 8K fixed the whole thing like magic. I haven't had a

>single snapshot too old in months and the RLB size is a mere 3Gb.

>