Quake II CTF server 16 player limit: PII 300MHz with 256 MB RAM and fast-ethernet LAN connection; server located in server room in a major Canadian university: measured packet rates using SNMP probes into server room core switch… less than 2% utilization of a fast ethernet connection, less than 500kb/s data. Those numbers were averaged across several dozen clan ladder matches, 8 ppl per team, each match was three games.
Also – LATENCY IS WEAKLY CORRELATED WITH ETHERNET LINE SPEED. Ethernet, fast ethernet and gig-E have increasing data transfer rates in the order listed. But any properly installed species of ethernet will have better than sub 10 mS latencies.
Latency is largely an architectural artifact. Ethernet will get you very roughly latency on the order of 1 mS. If you want order(s) of magnitude less latency you use different tech: infiniband or such.
Further: the most important network segments contributing to overall latency are the Internet itself and the endusers’ local loop (cable or DSL or dialup). The contribution of the server-to-hosting-company-external-router segment is (had better be) insignificant… and will be even if 10 Mb/s ethernet is used.
]]>N+1 CPU/Cores (minimum of ~1.9 ghz phenom I or equivalent)
(N*400mb)+300mb RAM
N = instances of dedicated server
Running a dedicated server on a single processor/core will open you up to lag spikes and stalls on the server side that will effect gameplay, as any service, network packet processing, or other background process can pause the processing of the game environment.
As for bandwidth, for a 8 player game the max needed connection is an upload of 1.1 Mbit/s on a very low latency connection.
A T1 connection at 1.536 Mbit/s > 100 Mbit/s cable connection.
Also, for good gameplay, you need to move out of your sisters house and dont use limewire. This will get you 3,400 more.
]]>