Re: Dead Space Map version 3 (simplified)

Hiroki Kataoka <kataoka(at)interwiz(dot)jp> wrote:

> Heikki Linnakangas wrote:
> > But does it
> > work for tables that have a small hot part that's updated very
> > frequently?
>
> I think there is no problem. Bloating will make pages including the
> unnecessary area which will not be accessed. Soon, those pages will be
> registered into DSM.

The usage_count of pages is always incremented when the page are accessed
even if they contain no useful tuples. Daed tuples in the pages are not
retrieved, but the pages themselves are considered as accessed.

Of cource, extra DSM-synchronizations are not needed if we assume that we
need to vacuum tables less frequently than checkpoints. However, we need
vacuum within 1min and set checkpoint_timeout as 15-60min in some cases.

Regards,
---
ITAGAKI Takahiro
NTT Open Source Software Center