[28984] in North American Network Operators' Group
Re: pop server in an ISP environment
daemon@ATHENA.MIT.EDU (Greg A. Woods)
Tue May 30 10:13:01 2000
MIME-Version: 1.0
Content-Type: text/plain; charset=us-ascii
Content-Transfer-Encoding: 7bit
From: woods@weird.com (Greg A. Woods)
To: nanog@merit.edu
In-Reply-To: <20000530185321.K79030@ewok.creative.net.au>
Reply-To: nanog@merit.edu (North America Network Operators Group Mailing List)
Message-Id: <20000530141046.31C02E0@proven.weird.com>
Date: Tue, 30 May 2000 10:10:46 -0400 (EDT)
Errors-To: owner-nanog-outgoing@merit.edu
[ On Tuesday, May 30, 2000 at 18:53:23 (+0800), Adrian Chadd wrote: ]
> Subject: Re: pop server in an ISP environment
>
> It depends entirely upon the useage. My little 386 gateway box could
> handle 20,000 pop3 users easily, but probably only a handful simultaneously.
> And then, it depends upon how big their mailboxes are, and what
> pop / mail software you are using, how you lay the mailboxes out on disk,
> how you've tweaked at OS you're running, whether someone spilt goats blood
> over the SCSI setup first, etc ...
>
> There really isn't a simple answer to this kind of question.
EXACTLY! ;-)
If those 20k users are all on permanent, fast, connections (cable, dsl,
LAN, whatever) and they all run something like Netscape with even it's
default parameters for checking for e-mail, and if those users have the
bad habit of leaving a bunch of mail on the server, then that dual-CPU
Sparc E250 will still fall to its knees if you used something
brain-damaged like even qpopper. However it could probably handle
200,000 users without too much trouble if you were using Cyrus-IMAPd.
I.e. if you're using system ids to authenticate and authorise users
(which is still the best and most secure way of doing things) then the
distinction of whether the system uses "short" or "int" (or "long", or
their unsigned variants) to represent the user-id internally is indeed
rather critical since we're crossing that 32k (or 65k) boundary quite
early in the growth curve.
(five "if"s -- definitely not a simple question to answer!)
--
Greg A. Woods
+1 416 218-0098 VE3TCP <gwoods@acm.org> <robohack!woods>
Planix, Inc. <woods@planix.com>; Secrets of the Weird <woods@weird.com>
