Packet Scheduling Algorithms to Support QOS in Networks.

Traditionally packet switching networks have supported only best-effort traffic. Newer applications need communication services that allow end clients to transport data with performance guarantees given in terms of delay, delay variation, bandwidth, and loss rate. The choice of the packet scheduling algorithm to be used at switching nodes is very crucial to provide the quality of service.

We have conducted a literature survey covering various scheduling disciplines that can be used to provide performance guarantees to clients. In this thesis, we have compared select scheduling disciplines in an experimental study using simulation. Weighed fair queueing, Class based queueing, and Rate-controlled static priority queueing, are the scheduling disciplines chosen for the study. The traffic types used in the study are CBR, VBR, ABR and UBR flows. For real-time flows used in the experiments, traffic source models for audio and video traffic were used as candidates for CBR and VBR flows. An extension of RCSP for using a new traffic model has been designed and experimented. We have also studied the use of RCSP for scheduling real-time flows in a link-sharing paradigm that uses CBQ.

Several experiments were conducted using the simulation test-bed for comparing the performance of the scheduling disciplines when serving various traffic mixes of CBR, VBR, ABR, and UBR flows. The conclusions from these experiments have been presented.