Re: XIDs and big boxes again ...

"Hans-Juergen Schoenig" <postgres(at)cybertec(dot)at> writes:

> i forgot to mention - i am on 8.1 here.
> so, VACUUM is not so smart yet.

So even if we added 64-bit xids it wouldn't be useful to you. You would have
to update (at which point you get all the other improvements which make it
less useful.) Or at the very least rebuild with the patch and dump and reload
which is just as hard.

> my changes are pretty much random I/O - so tuple header does not contribute to
> a lot more I/O as i have to read entire blocks anway.
> this is why i said - it is not that kind of an issue.

TPCC experiments show that even on random access you get the same performance
hit from bloat. I'm not entirely sure what the mechanism is unless it's simply
the cache hit rate being hurt by the wasted memory.

> and no, updating is not a 5 min task ...

I do hope you mean 8.1.11 btw. Updating your binaries should be a 5 minute job
and there are real bugs fixed in those releases.

--
Gregory Stark
EnterpriseDB http://www.enterprisedb.com
Get trained by Bruce Momjian - ask me about EnterpriseDB's PostgreSQL training!