[ut2003io] Server benchmarking?
pbruley at highstream.net
Mon Sep 30 21:24:06 EDT 2002
In my opinion BANDWIDTH is usually the biggest concern (assuming you are
not bogging your server down and the CPU is running at 1GHZ or higher)
Another Good Question would be: How much client-bandwidth per client is
required (client to server)and How much server bandwith is required
(server to all clients) and can these be controlled by the sever Admin?
What if a client is flooding the server with too much data how can the
server admin tweak the server to control the client side.
What rate and how many fps should the server serve up to the clients.
>There's been some discussions on server performance recently, and it
>seems to differ quite a bit from system to system, more then can be
>explained by just looking at raw cpu power, so I was wondering which
>features of a system have the biggest impact on server performance?
>Are UT2K3 servers in general more fpu or alu intensive?
>Does the memory need to be fast, or is a low latency the key to high
>performance? Does hyperthreading make a big difference when running 2
>servers on one cpu?
>Benchmarking is implemented very well for the client, but right now
>there's no good way to objectively test server performance with x
>players, which makes it hard for GSP's to decide what the best
>hardware is, or how many players a system should be able to handle.
>UT2003 has done a good job to become an established benchmark for
>gaming systems (smart marketing move that! :)), but the server seems
>to have been forgotten, even though it is an equally important factor
>in gaming pleasure.
>Would it be possible to implement benchmarking code for the server as
>well (internal demo playback or something?) or are the benefits not
>worth the trouble?
More information about the ut2003