Fair Traffic Scheduling and Shaping Algorithms for Diffserv Networks

Traditionally, the Internet has provided only best effort service to every user without any consideration to any requirements. As the Internet grows and becomes universally available, it becomes very important to deal with real time service delivery to application such as IP telephony. Thus, an interest has developed in having the Internet to provide some degree of Quality of Service (QoS). To provide different QoS commitments, the IETF developed many technologies that requires resources such as bandwidth and buffers to be explicitly reserved for a given data flow to ensure that the application receives its requested QoS. DiffServ is proposed by the thought that per-flow moved to the edge of the network with very simple functionalities left at the core network. However, services provided by DiffServ networks have lower flexibility and fairness among the aggregates sharing the network and among the flows of those aggregates. Per Hob Behavoiurs (PHBs) are implemented at DiffServ networks nodes using some scheduling and queuing mechanisms that are predecessor by markers and shaper. The current proposed DiffServ have unfairness problem that is caused by these elements. This thesis investigates the effect of different scheduling mechanisms and usage of a different marker algorithms then develop and implement a fair scheduling mechanism and integrate it with a shaper with the usage of a very fair efficient marker at the edge of the DiffServ networks with the usage of simulation and analytical techniques. In this thesis, I2tswTCM is proposed as an enhanced version of the famous tswTCM marker scheme used in the current DiffServ architecture, this new algorithm improves fairness in the excess bandwidth among different aggregates in a DiffServ networks and it has a better fairness than all other algorithms for different network provision levels. In addition, the algorithm marks the right amount of yellow traffic into the network proportional to the CIR. Nevertheless, it is not sensitive to the number of flows enters the networks. The thesis also proposed a new scheduling algorithm (FWFQ) which suite the DiffServ architecture and it gives a better performance in terms of delay and jitter. The research shows the importance of considering the scheduler type when dealing with different types of traffic sources in DiffServ networks. The interaction of different traffic type flows in a DiffServ networks using the proposed integration model between FWFQ and the dual token bucket shaper algorithm is analyzed. The result shows that the proposed model can significantly improve the performance.

Preview PDF 1600482.pdf Download (114kB)

Actions (login required)

View Item