Re: [Exim] Switching Over from Qmail for 20 millions mailing…

Pàgina inicial
Delete this message
Reply to this message
Autor: Nigel Metheringham
Data:  
A: exim-users
Assumpte: Re: [Exim] Switching Over from Qmail for 20 millions mailing load
On Wed, 2003-05-07 at 04:32, Nathan Ollerenshaw wrote:
> How is your hardware configured? Hardware has just as much a role to
> play in mail throughput as the MTA. For example, I've seen the
> throughput on sendmail quadruple just by putting it's inbound queue on
> a solid state disk. (Thats a battery backed ram-disk, basically).


This basically confirms the golden rule for MTA optimising. MTAs are
bottlenecked by disk I/O - especially synchronous disk I/O (when
receiving messages the RFC 821/2821 requirements mean that incoming data
must be written to stable storage before being acknowledged in the SMTP
session - this means both write and sync disk operations).

Ways round this are to make your disks faster (NVRAM or solid state
disks, or NVRAM journaled/cached conventional disks), or to just ignore
the requirements (leading to loss of mail under some circumstances).

    Nigel.
--
[ Nigel Metheringham           Nigel.Metheringham@??? ]
[ - Comments in this message are my own and not ITO opinion/policy - ]