Re: [glob2-devel] Finalized Design Idea for YOG

[Leo Wandersleb]

Stephane Magnenat wrote:
> > I'm missing a word on voip from your side. How would it fit into your
> > concept?
>
> Actually, voip fits fine into the normal order structure, I see no reason
> it wouldn't in the new one.

little traffic analysis. please correct me on the voip data stream that i 
considered to be 8kb/s and on the order size and frequency that i considered to 
be 200B every 2 s. (mousewheel might be many commands in one s. 200B might be 
far too much. I'd need some robust average)

voip:
big data stream per person that talks.
all voip data has to be routed to all other clients (n - 1).
in:  countOfPlayersTalking*8kb/s???
out: countOfPlayersTalking*(n-1)*8kb/s??
example (4 players; 8 players; 16 players) 3 talking [kb/s]:
in:  24; 24; 24
out: 72; 168; 360; (32 players: 744)
O(1) if never more than 3 are talking as this doesn't make sence. i guess there 
is no voice activation level implemented so all 3 talking might remain on their 
v-key producing a stream of data.

checksums:
constant stream of little data.
need no routing at all. they can be compared by the router.
size(checksum)=32b
in:  15*32b*n/s = 480b/s * n
out: 0/s (only in case of desync)
example (4 players; 8 players; 16 players) [kb/s]:
in:  1.9; 3.8; 7.5
out: 0; 0; 0
O(1)

zero orders:
size(zero_order)=80b?
in:  15*80b*n/s=1200b/s*n
out: 15*80b*n/s=1200b/s*n
example (4 players; 8 players; 16 players) [kb/s]:
in:  4.7; 9.4; 18.8
out: 4.7; 9.4; 18.8
O(1)

real orders:
they are rare like one order in two s.
they are small compared to voip.
size(order)=1600b??
have to be routed to all clients (n).
in:  0.5*1600b/s*n  =800b/s*n
out: 0.5*1600b/s*n*n=800b/s*n*n
example (4 players; 8 players; 16 players) [kb/s]:
in:  3.1; 6.3; 12.5; (32 players: 25)
out: 12.5; 50; 200;  (32 players: 800)
O(2)

putting it all together i see a problem with voip routing and with real big 
games where we have much routing of orders that is of O(2).

Greetings, Leo Wandersleb