Re: settings input for upgrade

Robert,

I was largely looking for input on whether I may have inadvertently shot myself in the foot with some of the choices I made when setting up postgresql 9.0, which is on different hardware than was the 7.4 setup.

The splitting of one table to two separate databases was done on 7.4 and did make a positive change in write performance. I was including that information only in an attempt to provide as much detail as possible.

- Midge
----- Original Message -----
From: Robert Klemme
To: Midge Brown
Cc: pgsql-performance(at)postgresql(dot)org
Sent: Saturday, August 20, 2011 2:38 AM
Subject: Re: [PERFORM] settings input for upgrade

On Thu, Aug 18, 2011 at 11:55 PM, Midge Brown <midgems(at)sbcglobal(dot)net> wrote:
> I'm in the process of upgrading from postgres 7.4.8 to 9.0.4 and wanted to
> run my decisions past some folks who can give me some input on whether my
> decisions make sense or not.

I am not sure what decisions you actually refer to here: in your
posting I can only see description of the current setup but no
decisions for the upgrade (i.e. changed parameters, other physical
layout etc.).

> The others are very write-heavy, started as one table within the original
> DB, and were split out on an odd/even id # in an effort to get better
> performance:

Did it pay off? I mean you planned to increase performance and did
this actually happen? Apart from reserving IO bandwidth (which you
achieved by placing data on different disks) you basically only added
reserved memory for each instance by separating them. Or are there
any other effects achieved by separating (like reduced lock contention
on some globally shared resource, distribution of CPU for logging)?

Kind regards

robert

--
remember.guy do |as, often| as.you_can - without end
http://blog.rubybestpractices.com/