Well, I agree with Suresh and Philip about the queue splitting is
generally a bad idea for Exim. Forget about it. I don't know about NFS
in modern kernels, in 2.2 it was a permanent headache BTW.
In such a nice datacenter produced by Suresh the main drawback becomes a
minor one: if a cluster member goes down, it takes down it's (small)
queue: this condition I actually meant by drawback. The smaller queue
means fewer problems.
Summarizing: large Exim installation should first minimize the resulting
`front end' queue size by any means. Queueing state, unfortunately, is a
wide-spread practice for ISPs, and I don't see mentionable trends to
address this problem by Exim itself. Spend money for the new shiny
clusters and smart heads.
2 Suresh: Thanks for the PPT.
-----Original Message-----
From: Philip Hazel [
mailto:ph10@cus.cam.ac.uk]
Sent: Tuesday, September 07, 2004 12:32 PM
To: Peter Savitch
Cc: exim-users@???
Subject: RE: [exim] Performance considerations
