Abstract
Because of the promised applications of Video On Demand, Video Conferencing and Distance Learning, video traffic is expected to account for a significant share of the total traffic volume in future ATM networks. MPEG (Motion Picture Experts Group), including I-Frame, P-Frame, and B-Frame, is the well accepted standard for video compression. In order to grarantee quality of service, it is important to police and manage the MPEG streams. In the ATM networks, Leaky Bucket algorithm is proposed to police the arriving traffic. When high priority cells disobey the traffic contract, they will be tagged as low priority. But in MPEG streams, I-Frame and P-Frame are essentially bursty, they will be tagged as low priority cells and lose to contend for the shared buffer. In this thesis, we propose a novel ‘Punish-Counter’algorithm. The high priority cells will not be tagged as low priority. Instead, those non-conformance cells will be counted for later punishment. When the shared buffer gets full, we select the low priority cells to replace according to their punish-counters. The simulation results indicate that our proposed algorithm can improve the cell loss ratio of I-Frame and P-Frame by 99% for single MPEG stream at high bandwidth utilization. Even in the case of the MPEG streams, we can decrease this ratio nearly to zero. Unless, the shared buffer is too small, our proposed algorithm can yield excellent performance in vaarious cases.