Date: Mon, 5 Dec 1994 12:55:02 -0600
From: cshotton@oac.hsc.uth.tmc.edu (Chuck Shotton)
To: Multiple recipients of list <machttp_talk@oac3.hsc.uth.tmc.edu>
Subject: Re: How to make the fastest MacHTTP server?

>What are the theoretical/practical limits on connections/minute in MacHTTP
>on various machines and how does that compare to a UNIX box?  This will be
>the big selling point or weakness of approaching the notion of MacHTTP for
>major servers.  Convenience and security aren't of great importance to
>companies who already have UNIX-happy personnel as it stands these days.

Macs will never compete head to head with high end workstations. It's an
apples and oranges comparison. MacTCP and NuBus ethernet cards cannot
provide the throughput to match something like a HP 9000/755 or bigger.
Currently, expect no more than 3 or 4 thousand connections per hour on even
the fastest Mac serving small docs on an ethernet.

When the Thread Manager version of MacHTTP comes out, I expect a large
performance increase, but it still won't match the beefiest Unix servers
one on one. Where the Mac makes up for this is in cost. You can buy 10
Power Mac 6100s for about $15,000, network them together, spread your web
pages across them, and destroy any $20,000 Unix server on the market. If
you set it all up correctly, the whole thing can easily be administered
from a single Mac. AND, you get incredible redundancy for free.

The mistake people make is to compare a single Mac to a single Unix host.
Compare the Unix host to a number of Macs that cost the same as the single
Unix host and the Macs will outperform the Unix host every time.

--_-_-_-_-_-_-_-_-_-_-_-_-_-_-_-_-_-_-_-_\_-_-_-_-_-_-_-_-_-_-_-_-_-_-_-
Chuck Shotton                             \
Assistant Director, Academic Computing     \   "Shut up and eat your
U. of Texas Health Science Center Houston   \    vegetables!!!"
cshotton@oac.hsc.uth.tmc.edu  (713) 794-5650 \
_-_-_-_-_-_-_-_-_-_-_-_-_-_-_-_-_-_-_-_-_-_-_-\-_-_-_-_-_-_-_-_-_-_-_-_-