Re: In praise of auto space management

In article <agkq46$gv5$1_at_lust.ihug.co.nz>, you said (and I quote):
>
> Do an insert. You get assigned to one of the free lists, and Lo! your new
> record is inserted into block 6 -because you got assigned to freelist 2.

Understood. But presumably (and this is a big *if*, here!), I'd expect that subsequent inserts would hit the first freelist - the one managing the first 5 blocks - as much as they hit the second one? So I'd end up at most with a small number of unused blocks in a large table, no?

Assuming a small number of freelists and I'm still on single user, mind you. Of course, concurrent inserts change the whole picture.

Is it round-robbin between the freelists, for inserts? Or is it "stick with one until it needs more extent, then go to next"? Never figured this one out to my entire satisfaction.

>
> In between the segment header, and the actual table data, there are 5
> completely empty blocks. Which all get scanned during Full Table scans.
> Hence freelists > 1 wastes space, and slows down FTS.

Cool. But this would only really be relevant for small tables and/or large number of freelists, no?

-- 
Cheers
Nuno Souto
nsouto_at_optushome.com.au.nospam