This example of voice traffic can be modeled with a simple two-state Markov chain. The chain consists of an ON state with exponentially distributed duration with average of 352 ms and an OFF state of 648 ms. In addition, the voice application is supposed to use the PCM encoding which generates 160-byte voice data every 20 ms.
TRAFFIC-GEN 1 2 DET 0S DET 600S RND DET 160 DET 20MS 0.352 NOLB
# {start} {duration} {traffic} {packet size} {interval} {prob} {bucket}
The entry in the application file specifies traffic from node 1 to node 2, beginning at time 0, with duration 600 seconds. The traffic is random (as opposed to file-based) with fixed size 160 bytes and fixed interval 20 milliseconds. The “prob” field is a value between 0 and 1 used to determine the percentage of the packets described by the given pattern that will be sent. This is to model the ON and OFF states (in a somewhat less than ideal way). In this case, only about 35% of the packets will be sent. If “prob” is set to 1.0, the traffic is essentially identical to the CBR traffic generated by “CBR 1 2 0 160 20MS 0S 600S.”
Note: There’s actually a slight difference because of how the applications schedule themselves.
The final parameter NOLB means that the traffic is not shaped by a leaky bucket.
The following graphic displays the result of this voice model. The throughput of the voice traffic and the overall average are shown. The overall average can be verified by 64 Kbps (= 160 bytes / 20 ms) * 0.352 = 23 Kbps.
This example is an arbitrary traffic model in which the data length is exponentially distributed with average of 160 bytes and the interarrival time is also exponentially distributed with average of 20 ms and the data generation probability is 1. There is no traffic shaping.
TRAFFIC-GEN 1 2 DET 0 DET 600 RND EXP 160 EXP 20MS 1 NOLB
# {start} {duration} {traffic} {packet size} {interval} {prob} {bucket}
The overall average rate of this traffic can be easily computed: avg. = 160 bytes / 20 ms = 64 Kbps. This average rate is displayed in the following Random Traffic Generation graphic.
