Some thoughts on how fgms bandwidth usage varies with #clients and #servers
-----------------------------------------------------------------------------

Anders Gidenstam


Definitions:

  B	client data production rate. [bps]
  n	total #clients
  n_i	#clients connected to server i
  N	#servers

Case 1

  Assume all clients are within range of each other. (E.g. if all are at KSFO)

  Received and sent data flows at server i:
    Source		Received	Sent				#
    one local client	B		B*((n_i - 1) + (N - 1))		n_i
    one remote client	B		B*n_i				n - n_i

    Total		B*n		B*(n*n_i + (N - 2)*n_i)		n

  Example: Data transmission rate of one server (B = 1)
    http://www.gidenstam.org/FlightGear/fgms/load_1,2,4_servers.png
           line 1: N=1, n_i=n		B*(n^2 - n)
           line 2: N=2, n_i=n/2		B*n^2/2
           line 3: N=4, n_i=n/4		B*(n^2/4 + n/2)


Case 2

  Assume none of the clients are within rage of any other.
  Assume the out of range inter-server relaying fraction is 1/10. 

  Received and sent data flows at server i:
    Source              Received        	Sent
    one local client	B			B/10*(N - 1)
    one remote client   B/10			0

    Total		B/10*(n + 9*n_i)	B/10*n_i*(N - 1)


  Example: Data transmission rate of one server (B = 1)

           line 1: N=1, n_i=n           0
           line 2: N=2, n_i=n/2         B*n/20
           line 3: N=4, n_i=n/4         B*3/40*n