I recall some list subscribers wanted to see some actual metrics for bandwidth utilization. Hope this is the sort of info you were looking for..... ---------------- Host: Dual xeon 2.8 Ghz 2GB RAM 100Mbps NIC (theplanet package labeled P2800S)
Running 8 srcds servers configured as follows: 1 8 slot scoutznknives 24*7 1 10 slot scoutznknives 24*7 1 12 slot awp_subzero 24*7 1 16 slot scoutznknives 24*7 1 24 slot cs_assault_lith 24*7 1 24 slot cs_assault_lith 24*7 1 24 slot various maps 1 32 slot cs_assault_lith 24*7 150 slots total
Current utilization: 53 players on line (player counts range from 10 to 70 over a typical 24 hour period with the average around 50) 1 server empty CPU is stable at 30% (+- 3%) across all 4 HT processors in task manager graph 575MB memory in use and stable CPU use rises to 60% in one HT CPU graph when a map change occurs
Network Bandwidth Utilization Metrics:
Current estimated transfer for March 2005 is 361.9 gigabytes (2000 gigabytes available)
Network utilization for past 24 hours: TO_INTERNET Max: 2.71 Mbits/s Avg: 931.8 Kbits/s Cur: 1.62 Mbits/s Total: 9.6 Gbits/s
FROM_INTERNET Max: 1.26 Mbits/s Avg: 460.8 Kbits/s Cur: 931.8 Kbits/s Total: 4.8 Gbits/s
TOTAL Max: 3.97 Mbits/s Avg: 1.36 Mbits/s Cur: 2.53 Mbits/s Total: 14.4 Gbits/s
Network utilization for past 72 hours: TO_INTERNET Max: 2.71 Mbits/s Avg: 839.7 Kbits/s Cur: 1.62 Mbits/s Total: 23.6 Gbits/s
FROM_INTERNET Max: 1.26 Mbits/s Avg: 409.6 Kbits/s Cur: 931.8 Kbits/s Total: 11.4 Gbits/s
TOTAL Max: 3.97 Mbits/s Avg: 1.22 Mbits/s Cur: 2.53 Mbits/s Total: 35.0 Gbits/s
Data transfer December 2004 was: 1,141.9GB. The daily average was around 36GB per day with a peak day on the 27th at 55.8GB and associated ramp up and ramp down during the adjacent 6 day periods. .On the peak day the average per hour was 2.33 gigabytes.
Although this might indicate player counts near 200 and the math on the numbers appear to indicate 184 players, I do not recall ever seeing more than 170 online at any one time.
But these are the actual recorded metrics that indicate I am not as full of BS as some folks around here suggest.
Summary
I'm not making any assumptions based on these numbers because I have some work to do on the log parser before I will have the necessary correlation of actual players online to the network metrics. When I can coorelate the actual players online to bits transferred then I can work it back through December 2004 and see what sort of player counts are suggested by the metrics and what the log parser shows for actual player counts and time online.
Other things I am keeping in mind is that a peak of 170 players for no more than 5 minutes would indicate the players were dropping off due to lag. However, if such a peak remained online and steady with the majority of players remaining in game for an hour then the indication would be the client FPS is satisfactory.
Also worth noting: There have been some network issues affecting this host that began in mid January with a malfuctioning Cisco router. The Fiber is scheduled to be regroomed tomorrow night. Perhaps this work will be successful and the player counts can return to the December levels by next month. -------------------------------------------
I'm still waiting for enough players to come online and take the CPU to a range of 80 to 90 percent (so i can enter the servers and measure fps/playability in game where it counts.)
IMHO, constant CPU utilization over 90% is not a good thing and you won't be able to tell what difference the numbers of servers and players is making. I am trying to capture the moment and counts when the server enters the 90% range so as to know how many players it took to get there.
More metrics later when I have them.
Regards...
_______________________________________________ To unsubscribe, edit your list preferences, or view the list archives, please visit: http://list.valvesoftware.com/mailman/listinfo/hlds
