[Date Prev][Date Next] [Thread Prev][Thread Next] [Date Index] [Thread Index]

(forw) [q2servers] Re: Best option for performance



Here's what I must have been thinking of. I must have confused the qw/q2
tests. I'll forward again if I here of a pure linux/NT test. 

Thanks for putting up with my  massive posts.


----- Forwarded message from "Carl E. Mankinen" <alginc@se.mediaone.net> -----

Received: from cornelius.INS.CWRU.Edu (root@cornelius.INS.CWRU.Edu [129.22.8.216])
	by b64198.student.cwru.edu (8.8.8/8.8.8/Debian/GNU) with ESMTP id SAA29903
	for <tod@b64198.STUDENT.CWRU.Edu>; Tue, 15 Sep 1998 18:04:40 -0400
Received: from list.criticalmass.com (list.criticalmass.com [204.182.161.114]) by cornelius.INS.CWRU.Edu with SMTP (8.8.8+cwru/CWRU-3.3)
	id SAA21521; Tue, 15 Sep 1998 18:04:38 -0400 (EDT) (from q2servers-admin@list.criticalmass.com for <tod@po.cwru.edu>)
Received: from 24.129.0.67 by list.criticalmass.com (Lyris SMTP service); 15 Sep 98 15:02:42 PDT7 from:<alginc@se.mediaone.net> to:<q2servers@list.criticalmass.com>
Received: from sphinx (alginc.se.mediaone.net [24.129.13.51] (may be forged))
	by mrout1.se.mediaone.net (8.8.8/8.8.8) with SMTP id SAA05710
	for <q2servers@list.criticalmass.com>; Tue, 15 Sep 1998 18:03:13 -0400 (EDT)
From: "Carl E. Mankinen" <alginc@se.mediaone.net>
To: "Quake 2 Servers" <q2servers@list.criticalmass.com>
Subject: [q2servers] Re: Best option for performance
Date: Tue, 15 Sep 1998 18:03:41 -0400
Message-ID: <177796-26734@list.criticalmass.com>
MIME-Version: 1.0
Content-Type: text/plain;
	charset="iso-8859-1"
Content-Transfer-Encoding: 7bit
X-Priority: 3 (Normal)
X-MSMail-Priority: Normal
In-Reply-To: <177769-6212@list.criticalmass.com>
X-MimeOLE: Produced By Microsoft MimeOLE V4.72.2106.4
Importance: Normal
X-Message-Id: <000001bde0f4$b2f4ed80$64641eac@sphinx>
List-Unsubscribe: <mailto:unsubscribe-q2servers@list.criticalmass.com?subject=[tod@po.cwru.edu]>
Reply-To: "Quake 2 Servers" <q2servers@list.criticalmass.com>
Sender: q2servers-admin@list.criticalmass.com
Precedence: bulk
X-Lyris-To: [tod@po.cwru.edu]
X-Lyris-MemberID: 26734
X-Lyris-MessageID: 177796

At my ISP, I tried NT4, Linux, and Solaris on several hardware platforms.
NT4, Linux were run on a 200MMX.
Solaris was run on Sun Ultra's, including a 5000e.

This was just prior to Quake2 being released, so all the testing was done
with QW and not with Q2.

Heapsizes were set to fixed at 16M if I remember correctly that the version
of QW at the time no longer supported the -heapsize cmd.

Tic rates were adjusted thru a wide variety of ranges, as well as rate
settings
and pushlatency.

We used a test group of 16 players using cablemodems with only 2 hops from
the
servers, and they were VERY low packetloss and low latency 50~100ms.

No matter what we tried on Solaris, it was slow.
It was a very strange sluggish feeling like you were pushing against
rubberbands
all the time. Nothing we tried could correct this. So we ditched the idea of
using Solaris.
Linux was better than Solaris.
However, NT4 was better than anything we tried.

NT4 was installed with a stripped down configuration and many
standard/default
services were disabled. No browser, server, netbios helper, alerter,
messenger,
ras, spooler, telephony, etc etc.
It was tuned pretty well, and it ran very much noticably better than any of
the unix versions we tried.
I don't know if this was the fault of the QW port on Unix, or if it was the
underlying OS that was causing the problems, and to be completely honest I
didn't
care because all that mattered in the final analysis was which one ran
better
for the clients.

Our Solaris guru was not able to stomach the results and so we spent
extensive time
trying to tune the solaris port to beat the NT performance.
This guy knows a LOT about Solaris too, since he used to be the regional
technical
rep for Solaris before he was hired by our ISP.

I was able to push ticrates much higher on NT4 and produce incredibly smooth
games.
Of course you had to have a pretty good connection to play.
You could toss a grenade and watch it arc thru the air without ANY jerkiness
or
missed frames or anything like that. It just looked smooth as glass.

So we went with NT4.
My current server handles about 150 players before it starts to get
unplayable.
This is because my cpu's are cooking, and I need to upgrade to faster
processors.

I am looking at the PentiumII Overdrive. If it can be run in a 4way SMP box
like
a proliant, then it might make a really good setup for running many more Q2
ports.


> -----Original Message-----
> From: q2servers-admin@list.criticalmass.com
> [mailto:q2servers-admin@list.criticalmass.com]On Behalf Of Rhomboid
> Goatcabin
> Sent: Tuesday, September 15, 1998 5:21 PM
> To: Quake 2 Servers
> Subject: [q2servers] Re: Best option for performance
>
>
> Do try the Intel pro/100+, I moved a news server and a web server, both
> running FreeBSD, to the pro/100+ from 3Com cards (a 905A and a 905B) and
> all my alignment, port, and packet requeues went away. The Intel card is a
> *great* card and I plan to upgrade all my servers to use them.
>
> If you can do a slim install of NT (no domain controller, no extra
> services, no internet info. server, etc.) you will probably see a little
> more performance than on Linux. Before I ran my FreeBSD vs. Linux tests on
> my 233 it was running NT 4.0 server. My problem was remote administration
> and the fact that I had installed all of NT's "server goo" mentioned
> above.  I'm building another machine and when I do I'll be doing a FreeBSD
> vs. NT comparison.
>


---
Brought to you by PlanetQuake: http://www.planetquake.com ... the epicenter of everything Quake.
Web Interface: http://list.criticalmass.com/scripts/lyris.pl?enter=q2servers
Browse/Search Old News, Edit Personal Info
To unsubscribe, forward this message to unsubscribe-q2servers@list.criticalmass.com

----- End forwarded message -----

-- 
Tod Detre |"If the women don't find you handsome, they should at
          | least find you handy." -Red Green 
          |"It is TOD not TODD! Do you see God spelling his name  
          | Godd?" -Me


Reply to: