Skip to Main Content
Traffic from real-time and multimedia interactive applications are known to be bursty. With this behavior, they are said to associate with self-similar property, by which they can no longer be expressed as Poisson distribution, but following power law distribution such as Pareto. Bursty traffic has a direct impact on network performance. Thus, if the self-similar property can be captured at earlier stage before submission to network, a specific mechanism could be applied at a sender node such that a more regulated traffic could be obtained. As a result, deterministic network performance could be attained, and thus allowing QoS guarantees be granted to users. However, this can only be effectively done with support of a source traffic analysis. Hence, the objectives of this paper are two folds: first is to identify the reason for burstiness, and second is to determine if the self-similar property can be removed from the source traffic. By executing these two, a traffic analysis shall be produced. Video traces of Jurassic Park and Soccer were simulated in a Bluetooth ad hoc network environment and were checked against a set of criteria for self-similarity. It was found that, in the first place, self-similar behavior is indeed associated with bursty traffic. Secondly, the number of packets produced from SAR segmentation protocol is the reason for the heavy-tailed distribution of the source traffic. Finally, the SAR protocol was found unable to eliminate the self-similar property from a bursty traffic flow.