On Mon, 2002-06-03 at 04:12, Marc MERLIN wrote: > On Sun, Jun 02, 2002 at 10:34:47PM -0400, Jon Carnes wrote: > > Lets not forget about the biggest bottle neck of all... the Hard Drive. > > For maximum through-put, you need to use a SCSI drive or RAID array as the > > Disk Subsystem for the server. If this is mission critical the obvious > > choice is a SCSI / RAID disk subsystem. > > Actually if you have enough RAM, the whole config.pck should fit in the disk > cache, so it's not a huge deal (but yes, I use SCSI and RAID for all my > servers)
Might be the case if your delivery system is different to your MLM system. Otherwise your MTA is going to do an awful lot of flushing stuff at disk during the SMTP conversations. [Does Mailman really not fsync its delivery data during message delivery?] Unless you are playing very fast and loose with the RFCs (821/2821) then you'll find that an MTA is strongly disk (transaction) limited. Hence discussions about journalling filesystems with NVRAM journals :-) Of course you probably *could* make a case for having less stringent disk requirements on an MLM - as long as you ensured a crash would only ever result in duplicate delivery - doing this could give you a significant speed up. Nigel. -- [ Nigel Metheringham [EMAIL PROTECTED] ] [ - Comments in this message are my own and not ITO opinion/policy - ] ------------------------------------------------------ Mailman-Users mailing list [EMAIL PROTECTED] http://mail.python.org/mailman/listinfo/mailman-users Mailman FAQ: http://www.python.org/cgi-bin/faqw-mm.py