| Microsoft DirectX 8.1 (Visual Basic) |
Determining the best configuration of transport topology, voice topology, transmission control, and codec depends on the type or genre of game you are creating, the number of players that will participate in a single game session, and the type of connection that will be targeted.
It's important to note that the number of players participating in the voice session is not necessarily the number of players actually participating in the game session. For example, if your game is a first-person shooter, voice communication can be represented in the game as a radio or communicator that is offered as a time-limited powerup. Also, the radio metaphor can be used to limit communication to radios in either vehicles or stationary command stations.
A second example to consider is an online bridge game, which involves four players at one time. Because this is a small working set, it is appropriate to choose a peer-to-peer voice topology transported over a peer-to-peer network topology. This small working set also allows for the use of voice activation as the mode of transmission control. The peer-to-peer voice topology is easily implemented and does not require any player to act as a server. If all four players use the Voxware SC6 codec, the maximum resulting bandwidth is 4.2 Kbps per speech stream, including the codec protocol overhead. Further assuming that game data requires negligible bandwidth, the outgoing maximum bandwidth requirement for an individual speaker is three independent streams to the other three players, or 12.6 Kbps. The incoming stream to any client ranges from 0 if no other players are talking, to 12.6 Kbps if all three other players speak simultaneously. The CPU requirement is 8 percent for encoding and 0 to 12 percent for decoding. This results in a worst-case requirement of 25.2 Kbps. Therefore each player must have a minimum of a 14,400-baud modem.
Another example is a squad combat game that can involve up to 32 players split between 2 teams. Assume that the game data requires a 28,800 baud modem. In this example, there is a larger number of players and it is appropriate to choose a forwarding server voice topology. Again, if all players use the Voxware SC6 codec, the bandwidth requirements are the same as the bridge game above: 4.2 Kbps. In this example we see that there is 4.2 Kbps outgoing when speaking, and a maximum of 12.6 Kbps incoming from the squad. The maximum CPU requirement is 8 percent of a Pentium 200 for encode and 12 percent receiving. Therefore, each player requires 28.8 Kbps for game data, and the greater incoming bandwidth of 12.6 Kbps requires a minimum 41,400 baud rate from each player's modem.
The worst-case scenario for the forwarding server itself is if all 32 players talk at once, requiring 134.4 Kbps. The server CPU use is minimal because the server is not encoding or decoding the streams. It is merely redirecting them. More typically, there might be 16 players talking simultaneously for 67.2 Kbps.
To illustrate the difference between choosing a mixing server voice topology and a forwarding server voice topology, consider the same 32-player squad combat game discussed above. If a mixing server voice topology is used, each client requires 4.2 Kbps to send and 4.2 Kbps to receive. The worst-case bandwidth requirements drop to 8.4 Kbps and 12 percent of the Pentium processor running at 200 MHz. This reduces the modem requirement to a 33,600 Kbps baud rate for the client.
For the server, the CPU burden changes. The server is now decoding and re-encoding all incoming streams and is also mixing the streams as required. The CPU burden mix the stream is relatively low and is considered negligible. The worst case is the decoding and encoding of 32 simultaneous streams. This results in a requirement of at least a Pentium II processor running at 400 MHz for the voice service alone.