[bf1942] CPU Usage

[Reinder P. Gerritsen]

> > Interesting facts there paul but just to correct you on one 
> > thing the HL engine is not one the lowest resource servers
> > by a long shot especially when running mods like CS / DOD. 
> > The best in my recent experience is either mohaa 
> > ( not mohaas ) and ut2k3 both of which have 
> > 1/1 player to cpu% relationship on our boxes. e.g. 20 
> > player mohaa = 20% this is in contrast to HL which always 
> > uses 1-3% even when empty and often peaks to 40% with 18 
> > players ( DOD mod ).
> >
> >     Steve / K
> 
> I ran the UT2k3 demo and early full version servers but found 
> it terrible in terms of stability, resources and 
> configuration, more trouble than their worth (plus the game 
> pretty much sucks imo). I dont know how you run your HL 
> servers but my ones sit pretty much at 0% CPU usage when 
> empty... this is all reletive to your CPU obviously and you 
> haven't mentioned what yours is in reference to your 
> percentages so your readings are pretty much worthless without this.
> 
> The Q3 based servers are pretty good with the server 
> resources although Spearhead uses a bit more load now, SoFII 
> is good with resources but uses a bit more than HL:CS. I 
> still say that the HL engine is still the most stable well 
> coded and optimised game server out there, this is probably 
> simply due to its popularity and therefore more resources 
> spent in developing it.
> 
> -Paul
> 

Have to agree with Paul here on part of the Unreal Tour. 2K 
UT2K is a disaster in configuration and updating (Linux version)
Web administration doesn't work correctly half the time, I made 
it once using the update tool @ instal time, neversince could be
updated again. In that way just as crappy as the Loki UT server 
release.
Stability and cpu usage wasn't much of a problem, mainly because 
there just were no players at all (most of the time).

Halflife as a good history of Linux server, originating from the
Good ol' QuakeWorld linux server. However, I've noticed some 
perculiar behaviour: servers running up to 12 players have 
relatively low CPU usage, 14 takes quite a step up, 16 players 
And up seem to take a leap into systems resources.
Another strange feature is that using "top", even though the first
Server is totally empty, it uses about twice the cpu time compared 
To the second instance, and the 2nd roughly twice to the 3rd.

(Running 6 servers,
2x 20 player DoD, 
2x 12 player Public CS, 
1x 14 player restricted, 
1x 20 player for wars.)

Also, to keep HL from acting unpredictable, I have a reset cron 
running at 6:30 am to kill off all servers and keep-alive scripts,
And have them started again at 7:00 am with a server start interval
every 5 minutes (having the 6th starting at 7:25, and the first 
server having its map changed at 7:30)

But all in all I'd say Quake derivative server engines have until now 
the best record of performance by far, making QuakeWorld, Halflife
And Quake3 relatively extremely wellplayed games on the net.
Lack of competition has kept these games in store for a looong time

I think BF1942 unlike RtcW has a real opportunity of taking over if 
the Linux server is becoming a succes in both performance, stability
and maintainability. Due to the bandwidth hungry nature of the game 
I think a stable Linux server will be the requirement to make BF, 
Even more then other games because there are to fiew Windows boxes 
on good connections.

Currently I'm planning on a P4/2,4 box to start a BF server on, but
I'm still a little worried about performance... Haven't been able 
yet to play a lot on internet, but unlike a 20 player DoD server 
which is  already loads of fun, I think 40 to 64 player games 
are somewhat of a must to make BF realy interesting.