Re: Suspected Postgres Datacorruption

> -----Original Message-----
> From: pgsql-performance-owner(at)postgresql(dot)org [mailto:pgsql-performance-
> owner(at)postgresql(dot)org] On Behalf Of Scott Marlowe
> Sent: Thursday, August 04, 2011 5:22 PM
> To: Sumeet Jauhar
> Cc: pgsql-performance(at)postgresql(dot)org
> Subject: Re: [PERFORM] Suspected Postgres Datacorruption
>
> > I ideally want to push to a higher version of Postgres . If I can
> prove that
> > there will be significant performance benefits and that crashes won't
> occur
> > then I will be able to present a strong case .
>
> Hehe. It would be hard to NOT get significant performance
> improvements moving from 7.4 to 9.0. Heck our load on our production
> servers went from 12 to 3 or so when we went from 8.1 to 8.3. Saved
> us a ton on what we would have had to spend to keep 8.1 happy.
> Install a test version of 9.0 on a laptop, point your test servers at
> it, and watch it outrun your production database for 90% of everything
> you do.

At a previous engagement, when we moved from 7.4 to 8.1 we saw a huge drop in transaction times. I don't remember the numbers but it was substantial. We also suffered very badly from checkpoint problems with 7.4, and we were able to tune them out in 8.1. When we went from 8.1 to 8.3, there wasn't an improvement in response times but we were able to deliver the same level of performance using a fraction of the I/O (due to HOT, autovacuum improvements the checkpoint smoothing stuff).

We also ran 7.4 for quite a while (on reliable hardware), and never had any corruption problems except for some index corruption issues - but that bug was pretty obscure and was fixed in 7.4

Brad.