OSEC

Neohapsis is currently accepting applications for employment. For more information, please visit our website www.neohapsis.com or email hr@neohapsis.com
 
From: Rafi Sadowsky (rafimeron.openu.ac.il)
Date: Sun Dec 02 2001 - 02:21:57 CST

  • Messages sorted by: [ date ] [ thread ] [ subject ] [ author ]

    On Thu, 29 Nov 2001, Wietse Venema wrote:

    >
    > With one recipient per message, disk I/O is the biggest problem,
    > and qmail does a bit more than Postfix. I suspect that's where most
    > of the speed difference comes from in his specific case.
    >
    > With small email messages, PIPELINING saves about 30% of the packet
    > count. The gain depends on number of recipients versus message size.
    > It's modest with one-recipient mail.

     Wouldn't that depend on link latency for total transaction time ?
    Compare for example a LAN link with ~1 msec RTT with a satellite link with
    ~650 msec RTT

    >
    > > In this case $smtp_destination_concurrency_limit would help a lot...
    >
    > Limiting concurrency while still keeping mail going is expensive,
    > memory wise; but it's part of being a good network citizen to not
    > make 1000+ parallel connections to the same site.

     Would that include Hotmail/Yahoo/AOL and similar that probably have load
    balancers behind every MX IP address ?

    >
    > Wietse
    > -
    >

    -- 
    	Rafi
    

    - To unsubscribe, send mail to majordomopostfix.org with content (not subject): unsubscribe postfix-users