Re: Dictionary-managed tablespace

Unless you use MINIMUM EXTENT, you'll find that Oracle rounds extent sizes up to a multiple of 5 blocks for a start. So after 32K, you expect 48K? But if you've got 8K blocks (just guessing!), 48K is 6 blocks, which isn't a multiple of 5. So it gives you 10 blocks instead -hence your 80K. However, I think it still remembers the 48K "ask", so the next request would be for 48+24=72K. Again, the nearest multiple is 80K. Next, you'll be asking for 72+36=108K. That's 13.5 blocks, so you get given 15, which is 120K. After that, you'll be asking for 108+54=162K, or 20.25 blocks. So you get 25 blocks instead, which gives you 200K.

And so on. My maths skills are sadly lacking, but you can work it out from there.

The usual caveats apply: PCTINCREASE should always and without exception be 0. Not 1, not 100. Zero. Every time. Based on the numbers you are reporting, you've got odd-sized extents in that tablespace, and that's a recipe for fragmentation.

Regards
HJR "EP" <ep_at_plusnet.pl> wrote in message news:agnee4$bgc$1_at_news.tpi.pl...
> Hi
>
> I'va got a question concerning dictionary managed tablespace - how are the
> extent sizes determined ?
> I normally use LMT so that issue hasn't troubled me so far.
> According to documentation, what matters here is INITIAL, NEXT AND
> PCTINCREASE.
> However, when I create a table with the following parameters:
> initial 32768 next 32768 pctincrease 50 I get odd extent sizes:
> 32768, 32768, 81920, 81920 (???), 122880, 204800 (???) 286720, 409600
> etc...
>
> TIA
> EP
>
>
>
>
>
Received on Fri Jul 12 2002 - 21:49:38 CDT