Owing to the existence of noticeable concentrated periods of contention and idleness, self-similar traffic can greatly increase packet delay and loss probability and thus reduce system resource utilisation. The development of efficient congestion control mechanisms plays a central role in the improvement of network quality of service (QoS), in particular for real-time multimedia applications. By exploiting the property of scale-invariant burstiness and correlation inherent in self-similar traffic, the authors propose an effective congestion control scheme, named adaptive wavelet and probability-based scheme (AWP), which concurrently operates over multiple time scales. AWP adopts the extended multifractal wavelet model (EMWM) for analysing estimated traffic volume across multiple time scales. Furthermore, a new auto-correction algorithm based on Bayes' theory for confidence analysis is employed to examine the validity of the predicted information. The analysis results can be used to enhance the adaptability of the prediction algorithm. In particular, the AWP framework can be easily extended to more than two time scales by increasing the level of wavelet transforms, which brings AWP a natural advantage in implementation and scalability. A series of simulation experiments have demonstrated that the proposed AWP scheme is superior to TCP and TFRC as it can greatly improve the QoS of multimedia data transmission while avoiding congestion collapse on the network.
Scopus Subject Areas
- Electrical and Electronic Engineering