From: | "Merlin Moncure" <merlin(dot)moncure(at)rcsonline(dot)com> |
---|---|
To: | "gnari" <gnari(at)simnet(dot)is> |
Cc: | <pgsql-performance(at)postgresql(dot)org> |
Subject: | Re: Hardware upgrade for a high-traffic database |
Date: | 2004-08-12 19:27:01 |
Message-ID: | 6EE64EF3AB31D5448D0007DD34EEB3412A7453@Herge.rcsinc.local |
Views: | Raw Message | Whole Thread | Download mbox | Resend email |
Thread: | |
Lists: | pgsql-performance |
> would one not have to repeat this operation regularly, to keep
> any advantage of this ? my impression was that this is a relatively
> heavy operation on a large table.
Yeah, it requires an exclusive lock and a table rebuild. It might be
useful to a message board type database since (one presumes) the reads
would be concentrated over recently created data, entered after the
cluster and losing any benefit.
As far as table size, bigger tables are a larger operation but take
longer to get all out of whack. Question is: what percentage of the
data turns over between maintenance periods? Plus, there has to be a
maintenance period...nobody does anything while the table is clustering.
Also, a particular method of reading the table has to really dominate as
far as user usage pattern. So, it's pretty rare to user cluster, but it
can help in some cases.
Merlin
From | Date | Subject | |
---|---|---|---|
Next Message | Arash Zaryoun | 2004-08-12 20:09:48 | Performance Problem With Postgresql! |
Previous Message | gnari | 2004-08-12 18:59:28 | Re: Hardware upgrade for a high-traffic database |