Re: Wrap-around XID's

On Tue, 8 Feb 2005, Alvaro Herrera wrote:

> On Wed, Feb 09, 2005 at 10:34:01AM +1100, Gavin Sherry wrote:
>
> Gavin,
>
> > Putting that to one side, the fact is that 32 bit XIDs only present a
> > problem to people who cannot vacuum their systems.
>
> This reminds me: how does Fujitsu-supported Postgres not need vacuum?
> Can you make any comment or description, beyond what's on the website?

I don't have any information on how Fujitsu's storage manager does this. I
presume they have some system to say that all current transactions see
this tuple as dead and then mark the space as reusable. It might be done
at the block level too. I do not have access to the source code so I can
only guess.

Another point worth noting is that they do have 64 bit XIDs.

Gavin