Priority Control in ATM Network for Multimedia Services

The communication network of the near future is going to be based on Asynchronous Transfer Mode (ATM) which has widely been accepted by equipment vendors and service providers. Statistical multiplexing technique, high transmission speed and multimedia services render traditional approaches to network protocol and control ineffective. The ATM technology is tailored to support data, voice and video traffic using a common 53 byte fixed length cell based format with connection oriented routing. Traffic sources in A TM network such as coded video and bulk data transfer are bursty. These sources generate cells at a near-peak rate during their active period and generate few cells during relatively long inactive period. Severe network congestion might occur as a consequence of this dynamic nature of bursty traffic. Even though Call Admission Control (CAC) is appropriately carried out for deciding acceptance of a new call, Quality of Service (QOS) may be beyond the requirement limits as bursty traffic are piled up. So, priority control, in which traffic stream are classified into several classes according to their QOS requirements and transferred according to their priorities, becomes an important research issue in ATM network. There are basically two kinds of priority management schemes: time priority scheme that gives higher priority to services requiring short delay time and the space priority scheme that gives high priority cells requiring small cell loss ratio. The possible drawbacks of these time and space priority schemes are the processing overhead required for monitoring cells for priority change, especially in the case of time priority schemes. Also, each arriving cell needs to be time stamped. The drawback of the space priority scheme lies in the fact that buffer management complexity increases when the buffer size becomes large because cell sequence preservation requires a more complicated buffer management logic. In this thesis, a Mixed Priority Queueing or MPQ scheme is proposed which includes three distinct strategies for priority control method -- buffer partitioning, allocation of cells into the buffer and service discipline. The MPQ scheme is, by nature, a non-fixed priority method in which delay times and loss probabilities of each service class are taken into account and both delay times and loss probabilities can be controlled with less dependency compared with the fixed priority method, where priority grant rule is fixed according to the service class, and the priority is always given to the highest class cell among cells existing in the buffer. The proposed priority control is executed independently at each switching node as a local buffer management. Buffer partitioning is applied to overcome the weakness of the single buffer