Re: [NEWBIE] CircleMUD and problems with other programs

From: Jörgen \ (di5sig@CSE.HKS.SE)
Date: 07/30/97

> Granted a *nix box may be more suitable for running a mud, a NT machine
> is no more subject to cpu or memory lockup than a *nix box. It all depends on
> the priority given to the process and the effectiveness of the programming
> behind the mud. Give a big mud a priority of 20, and you can see a *nix box
> lock up pretty quick also.

Whatever the demandpaging/scheduling algorithms Microsoft use in NT, they do
not work properly. I've seen Sparc-5's running Solaris 2.5 with only 32 Meg
memory that runs ALOT better than P200 with 32 Meg memory running NT 4.0.
(FYI, the sparc 5 has a supersparc 75 Mhz cpu if I'm not wrong, and that cpu
is a lot less powerfull than a P200). Nice CPU-utilization on the Solaris,
sucky on the NT.

NT machines can be deadlocked (almost deadlock, its a pretty similar shitty
state) very easy. Too easy indeed. A Unix machine however can handle priorities
a lot better.. I've managed to boost a P166 up to 25 in load and still being
able to stop it (Linux). For those who don't know, 2 - 3 is a lot in load..
I used recursive forks and mallocs for this, which is very painful for the
kernel. A mud doesnt do this, unless the coder has this brilliant scheme in the
back of his head. ?] (well it does allocate memory.. but not in like 100
processes at the same time :) So, it is VERY unlikely that circlemud (or any
other mud for that matter) will lock up a UN*X machine (The star goes between
N and X).

Trust me.. I've been there, done that.. Might be doing it tomorrow.

// JOrgen

     | Ensure that you have read the CircleMUD Mailing List FAQ:  |
     | |

This archive was generated by hypermail 2b30 : 12/08/00 PST