Coding for 32 Players

[malric]

A bit more on topic. A find it strange that the bandwidth is n^2. Shouldn't it be a linear function ?

n clients making orders, n clients to send them to

n*n

n^2

My mistake, I read the first post something like "for a client the host will send 2^n". You meant, the total amout, not the amount per client.

Anyhow from this topic:
http://spring.clan-sy.com/phpbb/viewtop ... h&start=20
someone said per player as host it used 1.5k - 2k per player. So I think at least some people should be able to host a big game.

[REVENGE]

Ouch...that would mean 4096kbs of bandwidth necessary.

[smoth]

your post breaks the thread please fix it

[REVENGE]

your post breaks the thread please fix it

Hmm, fail post fails thread.

[bibim]

For information, here is an extract of the bandwidth usage graph of DediServ today (it was 5v5 and 6v6 games):


But you need much more bandwidth than what is really used to offer a good quality of service (unless you set up specific QoS rules of course...).

[SwiftSpear]

woah... I would have thought that bandwidth usage would increase as the game went on as opposed to slowing down...

Also, is there any explanation of the games where the upload was so significantly larger than the download? Possibly lots of specs? Not that it's massively significant, but it would be neat to be able to figure out exactly what effect extra spectators have on a server (since they get blamed for lag all the time )

[Tobi]

woah... I would have thought that bandwidth usage would increase as the game went on as opposed to slowing down...

Maybe this is because less micro happens (less CMDs/min) late game and/or maybe some players dropped? (This effect would need to be stronger then the increase in number of units per command.)

Also, is there any explanation of the games where the upload was so significantly larger than the download? Possibly lots of specs? Not that it's massively significant, but it would be neat to be able to figure out exactly what effect extra spectators have on a server (since they get blamed for lag all the time )

I guess that could be specs yeah; after all they don't give orders so only traffic is for chat/ping and to send players' commands to them.
(ie. adding a player has the effect of increasing download of everyone AND upload of host, adding a spec has only the effect of increasing upload of host.)

[LordMatt]

I think that graph is more than one game.

[Tobi]

Yes, but I think the individual hills are single games.

[bibim]

Yes, there are 8 games on this graph. The game starting at 17:26 was a game with one spec at start (during the game some players became spec also).