Re: [NEWBIE] CircleMUD and problems with other programs

From: Gary Barnett (gbarnett@POLARNET.COM)
Date: 07/29/97


On Tuesday, July 29, 1997 12:09 PM, Patrick J. Dughi [SMTP:dughi@IMAXX.NET]
wrote:
> > > causing problems is how does CircleMUD interfere with web servers, mail
> > > servers, and other communications programs that are running?  On a
> > > Windows
> > > NT 4.0 server running multiple programs would MUD crashes and general
> > > MUD
> >  It shouldn't... They say that NT's are smart enough to prevent one
> > program from crashing whole system.
>         This is true. However, like windows95, one process can lock up the
> entire machine - stealing too many cycles, and memory...however, comparing
> windows95 multitasking to NT multitasking is like the ELF to an AppleIIe.
> Yeah, they both suck, but at least you can play "Moonpatrol" on an
> AppleIIe

Let's clear this up a bit with facts, shall we?

1) Win 32 (95/NT) when running 32 bit apps, is capable of memory protection
and preemptive
    multitasking -- much as unix does, one process is not capable of stealing
more cpu cycles
    than the OS is willing to give it.

2)  Win16 (Windows 3.1 and all windows 3.x apps) The OS uses cooperative
multitasking. NT
    runs a seperate virtual machine for 3.x apps and runs them all under this
process. One 3.1
    app can hang the entire 3.1 VM, but, unless the small amount of
non-protected memory is
     trashed, the win32 apps will continue running. This is a compatiblity
mode.. and on the way
     out.. so if you want to compare NT to Unix, do it w/o running 16 bit
apps in order to get an
     apples to apples comparison.

3) NT handles memory protection better than 95, due to 95's reliance on real
mode MSDOS
    calls. Therefore, in the same crash, NT stands a better chance of not
being affected. NT
    does not handle memory protection as well as Unix. There is a small area
of memory
    that can be trashed. There are also tools to help you determine why and
who...So.. there's
    the main reason all these threads go on and on. Unix protects it's memory
better than NT.
    The usability of NT as a mud server depends on those who write the code
and their QA
    procedures. Running a buggy mud under NT is going to cause more OS
crashes than Unix.
    Running a production mud under NT probably won't.. fewer bugs, the small
memory chunk
    fades to the background, as it's no longer a significant issue in server
availability.

>
> >
> > > use cause problems with the other programs or cause them to crash as
> > > well?
> > > I have heard all about MUD crashes but I haven't actually seen one and
> > > I
> > > don't know what it does to the system.  Also, with 32 mb ram and a
> > > pentium
> > On Linux, you get a nice core file... And that's about it... :)
>         A mud crash is simply the actual program dying. Usually it dosen't
> affect the rest of the system.
>
On Unix you can get several types of errors that are fatal to the system.
With Unix
(as well as NT) these are usually hardware configuration errors, bad drivers
or
faulty OS code.

> > > pro 200 would the system be slowed down by the MUD?  I would hate for
> > > his
> > WindowsNT running on machine with 32 megs of ram? *shiver* Brrrr... Mud
> > will slow it down a bit... but since it's already like a snail it won't
> > show much ;>
>
A mud does very little disk IO, and unless it's code and data gets paged out
a lack of
physical RAM, there will be no speed hit, due to this. The act of other
programs paging
in and out will affect response time, especially if you use an IDE drive. If
you want a fast
server, use SCSI; the processor will escape the responsibility of
baby-sitting the drive.
This holds true for Unix, just as it does for NT.

>         Yes. NT, with 48 megs - running only the microsoft backoffice
> webserver, with frontpage extensions causes the screen saver to become
> jerky, much less allow you to run other programs. It would very much slow
> it down.  Of course, the NT machine i'm refering to is only a p133, but
> i've heard tell that NT isn't really happy till you get about 64 megs.

NT (in a usuable configuration) uses much more than the 12 megs MS would
have you believe to run basic OS functions. 64 is a pratical lower limit for
doing
anything in the background (Like back office) and not experience excessive
page ins/outs, which slow the system to a crawl.  BTW Running a screen saver
is a bad idea on a file server. If you must run something, run the blank
screen one.

We run 130 megs of ram in our Winframe server (multi-user NT) and 64 megs in
the
RAS, NT Server and SNA servers I support. Even as a desktop OS, paging
becomes
a problem with less than 48 megs of ram with NT.

>
>         I really recommend that you gauge the system responsiveness after
> running your mud, if you're worried.  Lucky for you that most people web
> browsing expect a delay when they connect.. modem users for the most part.

Stock circle uses about 6 megs of memory if memory serves. Use of the system
monitoring
tools would be a good way to determine how much memory is being allocated
during normal
use. Then add 6 megs to that, add 50% for disk caching and compute the memory
 you
need to add to the system.

A mud uses very little processor time or bandwidth. There should be no reason
why an NT
server, properly configured, should yield aceess times over 500msecs
(including 300 msecs
for network travel time.) That's including some latency for other tasks
operating and using
the drive (the slowest component) plus latency from modem access.

--Mallory


"Never worry about the bullet with your name on it. Instead, worry about
shrapnel addressed to 'occupant.'" --Murphy


     +------------------------------------------------------------+
     | Ensure that you have read the CircleMUD Mailing List FAQ:  |
     | http://democracy.queensu.ca/~fletcher/Circle/list-faq.html |
     +------------------------------------------------------------+



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