Re: Postgres memory question

In response to Kobus Wolvaardt <kobuswolf(at)gmail(dot)com>:

> 2009/8/9 Scott Marlowe <scott(dot)marlowe(at)gmail(dot)com>
>
> > On Sun, Aug 9, 2009 at 4:06 AM, Kobus Wolvaardt<kobuswolf(at)gmail(dot)com>
> > wrote:
> > > Hi,
> > >
> > > We have software deployed on our network that need postgres, we have
> > server
> > > that hosts the server and all worked fine until we crossed about 200
> > users.
> > > The application is written so that it makes a connection right at the
> > start
> > > and keeps it alive for the duration of the app. The app is written in
> > > Delphi. The postgres server runs on a windows 2008 server with quad core
> > cpu
> > > and 4 GB of ram.
> >
> > Is this an app you can fix yourself, or are you stuck with this
> > mis-step in design?
>
>
> It is our app but it is not going to be easy to change it. It will get
> changed, but the time frame is a bit long and we need a solution to hold us
> over.

We have servers using about 200 connections on average ... it climbs up
to 300+ during busy use. I've seen it peak as high as 450, and we've seen
no performance issues.

This is a quad core with 4G of RAM. Of course the OS isn't windows, it's
64 bit FreeBSD.

However, my point is that your server _can_ be tuned to handle loads like
this. I can't say for sure how much the OS makes a difference in this case,
but I expect any POSIX system is going to scale better than Windows.

As far as tuning, I just went through the config file and tuned everything
logically based on published best practices. Aside from the FSM settings,
I don't think I've had to fine tune anything else, post.

And for those who may want to jump in -- we have investigated pgpool several
times, we just can justify the added complexity when the system just works
as is, but we're ready to add it on quickly should problems arise.

--
Bill Moran
http://www.potentialtech.com
http://people.collaborativefusion.com/~wmoran/