All three servers are running pretty much wide open.
HLDM:
sv_maxrate 9999
sv_maxupdaterate 30
HLDM:S:
sv_maxrate 20000
sv_maxupdaterate 60 (recently lowered to 40 though)
HL2DM:
sv_maxrate 20000
sv_maxupdaterate 60 (might lower this to 40)
Note: Most HL2 clients have the default cl_updaterate 20 and rate 20000 set.
The total number of players (62) times the total number of max updates per
second (150) give a possible 9,300 possible updates per second.
On a wide-open client I see high-end incoming data at (all bytes / sec):
hl2dm:
20k/s incoming on the client at 20000 & 60/s settings (has physics
data) (server typically reports 154k/15 players outgoing - 10k/player)
hldm:s:
12k/s incoming on the client at 20000 & 60/s settings (has minimal
physics data) (server typically reports 60k/7 players - 10k/player)
hldm:
5k/s typically on the client at 9999 & 30/s settings
so to calculate ~maximums from the client side (all bytes/s):
hl2dm: 20k/s per client * 16 clients = 320K/s outgoing max bandwidth from
the server
hldm:s: 12k/s per client * 32 clients = 384K/s outgoing max bandwidth from
the server
hldm: 5k/s per client * 14 clients = 70K/s outgoing max bandwidth from
the server
------------------------------
total: 774KB/s ~= 7.74Mbits/second estimated maximum network demand
Looking at the network performance on taskmanager, I'm currently seeing
peaks of 2.5% of the 100Mbit/sec connection or 2.5Mbit/s outgoing, but all
three servers are usualy not at 100% (hldm:s is typicaly not going over 20
players right now).
A crosscheck against the monthly totals that are being reported:
say I'm getting a daily average of 2% of the 100Mbit/s network, so 2mbit/s
sec
2mbit/s * 3600s/h*24h/d * 30 days = 5184Gbits/month / 10bit/byte =
518.4GBytes/month which is almost exactly what I'm seeing reported as usage
right now.
It would appear that the individual client reporting is tracking reasonably
well with the server side totals, with the raw network measurements and with
the actual recorded usage. Actual server usage measurements are probably
read directly from the local network switch (they only seem to care about
outgoing usage).
What I don't really see are the instantaneous peaks (things are all averaged
somewhat in these measurements).
This is all a tremendous load for a single 3.0GHz CPU, I can easily see a
need for a real dual processor system here.
The good news for memory usage though is that all three servers are only
using about 580MB out of the the 2GB on the server box.
qUiCkSiLvEr
----- Original Message -----
From: "Whisper"
To: <hlds@list.valvesoftware.com>
Sent: Sunday, September 24, 2006 6:34 PM
Subject: Re: [hlds] Server "stats" report interesting note (HT related) &
server total bandwidth usage
What were the servers rate settings?
On 9/25/06, Kevin Ottalini wrote:
Since I've never seen a factual report of actual server bandwidth usage,
I thought this might be interesting to everyone.
I have three servers running on an Intel platform under Win Server2003,
Hyperthreading is enabled.
When I ran "stats" on the servers when they were very busy, I saw two
servers reporting 75% CPU each and one server reporting 30% CPU - that
adds
up to a total of 180% CPU utilization. The reported usage matches what
was being shown graphically in the server GUIs.
Task Manager was reporting about 90% utilization during that time, so it
makes sense that the system is acting as if there were two processors
available.
I did some testing earlier and there was very little performance
difference with HT enabled or disabled, although the server in general
seemed slightly more responsive with HT enabled.
Bandwidth usage - The three servers have been running a month now and
have been very busy.
There is a 32-player server, a 16 player server and a 14 player server so
max 62 players, but the max has been about 50 players on at the same
time.
(I think the 32 player server is overloaded, probably needs to be limited
to no more then 20 players).
The platform has been extremely stable with no crashes, the only time
I've had to reboot was to install OS updates. I run no addons on the two
Source servers and I only run metamod and stripper2 on the HLDS server.
The game server and corresponding web server utilization are:
Server box (3 game servers):
Bandwidth used: 388 GB of allowed 500GB/Month in 22 days, 30 day
total/month = 17.64GB/day = 529GB/month (I need to keep an eye on this
hehe)
webhosting (fast http):
Bandwidth: 21.577.04 GB of allowed 2000.0 GB/month in 22 Days so about
1GB/day (includes the server web page).
qUiCkSiLvEr
_______________________________________________
To unsubscribe, edit your list preferences, or view the list archives, please
visit:
http://list.valvesoftware.com/mailman/listinfo/hlds
