Re: Pre-Allocate table size with minextents

Contiguity of extents is a complete and utter myth, and there is no benefit whatsoever in attempting to achieve it.

Don't specify minextents. At least, don't specify such a large minextents that you've chewed up two years' space. Or if you must do that, at least do what the paper says and stop worrying about it and start living. (That is, so long as all your tables have two years'-worth of growth in hand, why worry about the lack of free space within the tablespace?)

If your tablespaces are locally managed, then the cost of acquiring an extra extent when needed is pretty low and the benefits of pre-allocating the space are accordingly pretty limited.

Regards
HJR "Philippe LAVIGERIE" <plavigerie_at_transiciel.com> wrote in message news:arb5sd$9ug$1_at_s1.read.news.oleane.net...
> Hi ,
> Before creating my objects, I follow the advises of the document 'How to
> stop defragmenting and start living: The definitive word on
fragmentation'.
> So, I have 3 tablespace initial extent size : 128Ko, 4Mo and rarely 128Mo.
> Based on the number of lines I should have at the beginning of the
project,
> the lines added every month, and the number of months I should keep
online,
> I make a table size estimation.
> Then, I create the table with a MINEXTENT corresponding to the estimated
> size. So, I will 'reach' this size after (for example) two years. I know
> that if my estimations are bad, I can waste a lot of space.
> As I can have a lot of tables in the same tablespace, is it better to
> pre-allocate the space for a table (meaning that all extents are
contigous)
> or to leave MINEXTENTS to 1 and the extents will be sparsed in the
> tablespace ?
> Could you share your experience with me ?
>
> P.S : when I pre-allocate the table, I must use DBMS_SPACE to see the free
> blocks because the tablespace map indicate that there is no space left.
>
>
>
Received on Mon Nov 18 2002 - 13:12:42 CST