Return to BSD News archive
Path: euryale.cc.adfa.oz.au!newshost.carno.net.au!harbinger.cc.monash.edu.au!munnari.OZ.AU!news.ecn.uoknor.edu!feed1.news.erols.com!howland.erols.net!newsfeed.nacamar.de!news1.best.com!nntp1.ba.best.com!not-for-mail
From: dillon@flea.best.net (Matt Dillon)
Newsgroups: comp.unix.bsd.freebsd.misc
Subject: Re: Network Performance
Date: 3 Jul 1997 14:47:29 -0700
Organization: Best Internet Communications, Inc. - 415 964 BEST
Lines: 112
Message-ID: <5ph6lh$3ee$1@flea.best.net>
References: <slrn5qngmn.ie2.jlc@mass-pc.wpi.edu> <5opk40$aho@nusakan.cedar.buffalo.edu> <33bac1aa.215140895@news.probe.net> <33BB878E.15FB7483@rockhead.com>
NNTP-Posting-Host: flea.best.net
Xref: euryale.cc.adfa.oz.au comp.unix.bsd.freebsd.misc:43882
:>PC's aren't necessarily "cheap" these days but, why not run on the least
:>expensive platform that will do the job? And what does FreeBSD run on?
:>Isn't it the same cheap PC as NT?
PC's *ARE* cheap compared to vendor-specific boxes (SGI, SUN, HP,
DEC, etc...). And once the Pentium Pro came out with its memory ECC
capabilities, it was all over for those boxes. The equivalent
hardware using PC technology costs between 1/3 and 1/2 what the same
hardware costs using vendor-specific technology.
What's left for the vendor-specific boxes are the higher-end
multi-processor 'enterprise server' boxes. In my view, it's
a lost cause, because the major cost these days is NOT the
core box but is, instead, the price of memory and disk. Why
pay a 200% to 500% premium for SMP when you don't need it ?
This leaves us with three players in the game:
(1) FreeBSD/NetBSD/Linux
(2) Solaris (running on PCs)
(3) NT
:>At this point you can surely see the need for a "goals" document that
:>specifies why you are evaluating NT and what your systems need to be
:>doing.
:>
:> How many simultaneous connections,
:> How many packets per second,
:> How many bits per second,
:> Reliability, blah, blah, blah, maintainability, blah, blah, blah
I think it goes beyond simple statistics. Microsoft can
fake most statistics to make NT look good even when it isn't..
Do you remember the database 'clustering' demonstration they made?
it was pathetic... they mangled the definition of a 'cluster' so badly
that it wasn't even close to reality, yet they managed to convince
half the people at the demonstration that what they had was useable.
:>Set forth some hard definable mesuarements before hand, other wise the
:>suits will massage the data to fit the decision they've all ready made.
:>I'm not talking numbers but, things to be measured. Then compare the
:>empirical data.
:>
:>What's the most likely outcome? When the suits start talking "let's
:>try NT you can't go wrong with M$" your suit can pop out this little
:>bit of info and say "we tried that, it has some problems",
What we need to be is more vocal. Whenever someone brings up
some stupid, mangled proprietary feature introduced by Microsoft,
we need to step up to the plate and shoot it down: "You must be
joking, this would lock us into a proprietary protocol and single
vendor when we could instead use XXXX, which is a world-wide
standard and has several vendors in competition".
Or if, as they so often do, microsoft introduces some incremental
feature that nobody else has, you have to take the tack: "This
is something microsoft threw together in a few weeks to try to
lock people into using only their products. We don't even NEED
this feature, why the hell are we considering it?".
One can always point to various disasters, e.g.: It took
microsoft two years to get FrontPage to where it is today,
and it's STILL horribly broken... are you seriously suggesting
that we hop on board this other product YYY which microsoft
has only just released? Why should we throw away good money
alpha-testing microsoft software ?
Or: "NT worked great for me until something broke, then I was
up shit creek".
My favorite... one that I hope to leverage myself: Struggling
company A points to healthy company B and says "Look at them: they
are using UNIX on all of their platforms and they haven't had any major
blowups or problems. How the hell are we supposed to compete
with them using this NT shit?". Or: "Look at them: they are
doing their entire service on just 4 machines. We are spending
tens or hundreds of thousands of dollars on 30+ machines running
NT and can only support half the users they do!".
I want BEST to be one of those 'healthy companies running UNIX
and beating the shit out of any competition using NT'. I want
paste the words right in front of the faces of the IS departments
of other companies: "Look, we are bringing up servers at 1/6
your cost. Only an idiot pays $5000 for operating system software
which doesn't even come with real support. Microsoft support?
Yah, right. We tried microsoft support... they know less about
multi-user/server systems then we do! They fix bugs by calling
them 'features' and forcing the rest of the world to conform
to them. No thanks!".
This is my declaration:
We (BEST) now have 25 rack mount ppro 200's running FreeBSD UNIX,
doing mail, web hosting, news, shell, and other services. None of them
are overloaded. All of them are highly responsive, all of them are
extremely reliable, and each user box is able to support over 2000 accounts
without even breaking a sweat. They don't crash, they don't degrade over
time, There are no weird quirks when running multiple services on
a single box, they don't loose track of memory or other resources
over time, the operational maintenance is trivial (one good person
could basically maintain ALL of the boxes), the machines are
highly secure, resist internet attacks ten times better then
NT ever could, and we get a continuous and timely stream of
software updates. There are no licencing issues, and no worries.
Only a fool would try to compete with us using an NT base to
run his services... probably the same fool who is willing to hire
'NT' professionals bearing certificates of training from community
colleges or extension courses!
-Matt