Path: utzoo!utgpu!news-server.csri.toronto.edu!rpi!usc!samsung!uunet!panews.awdpa.ibm.com!qe2.awdpa.ibm.com!steve
From: steve@panews (Steve DeJarnett)
Newsgroups: comp.mail.misc
Subject: Re: mail gateways
Message-ID: <1991Mar20.175058.24212@ibmpa.awdpa.ibm.com>
Date: 20 Mar 91 17:50:58 GMT
References: <1991Jan22.053940.24424@pa.dec.com> <1240002@hpcc01.HP.COM>
Sender: news@ibmpa.awdpa.ibm.com (news id)
Reply-To: steve@qe2.awdpa.ibm.com (Steve DeJarnett)
Organization: IBM AWD Development, Palo Alto
Lines: 55

In article <1240002@hpcc01.HP.COM> cricket@hpcc01.HP.COM (Cricket Liu) writes:
>Brian,
>I run HP's primary mail gateway, relay.hp.com.  I'd be happy to compare
>notes with you.

	OK, I don't actually even run our gateway, but here's the stats off of
it.  Note that this is NOT ibm.com.  This is simply our site's Internet 
gateway, which really only passes mail for user's here (about 65) and at a
couple of other sites within IBM.

>    For example:  our typical load on a business day is 10,000 messages
>    which get delivered to 17,000 recipients.  Is that a lot or a little?
>
>That sounds like a lot.  I'm just starting to get good numbers after our
>syslog-muncher broke, but after a quick look at the last hundred or so logs
>(ugh), I'd say we average about 4000 messages per day, with a peak of about
>10000.  I'm afraid I don't have figures for the number of recipients.

	Well, our average is about 775 messages a day, with peaks near 1000.
No recipient figures here either.

>    What hardware do you use?  We use a pair of DECsystem 5400 computers,
>    which are about 25 MIPS each, and which have 64 MBytes of memory.
>
>relay is an HP9000/845 (about 24 MIPS) with 48 MBytes of memory.

	We use an IBM RT Model 135 running BSD4.3 (maybe 6 MIPS, downhill,
with a tailwind :-) with 8 MBytes of memory (well, it's not that busy, so it
really doesn't need more).

>    That says that our crisis peak message-handling capacity is 22,000
>    messages per 3200 MIP-megabytes, using speed cross VM occupancy is a
>    crude measure of capacity.  That's about 7 messages per day per MIP
>    megabyte.  Is that a lot or a little?
>
>Well, if I take 10,000 messages per day to be relay's peak, then our max
>message-handling capacity is ~6.5 messages per day per MIP-megabyte.
>Sounds like we're right in the ballpark.

	OK, if we take 1,000 messages per day to be ibminet's peak, then our
max message-handling capacity is ~20 messages per day per MIP-megabyte.  Then
again, we're not really a major gateway, and our message delivery time has been
quite reasonable, so I don't consider this to be too far out of line.  The
sample size here is too small to merit much consideration, anyway.

>Now I wonder if IBM will give us their stats...

	Now you know...:-)  A more interesting set of stats might be to look at
ibm.com or the iinus1.ibm.com gateway to see what kind of traffic they're 
seeing.  However, I doubt anyone is going to go out of their way to get mail
stats for a machine running CMS...:-)

	Well, that was educational, if nothing else.  Better late than never...

	Steve