|M.Sc Student||Dvorkin Alexander|
|Subject||Ensuring End-to-End Delay in the DiffServ Model|
|Department||Department of Computer Science||Supervisor||Professor Dan Raz|
Quality of service (QoS) is an umbrella term for a collection of technologies that allow network-aware applications to request and receive predictable service levels in terms of data throughput capacity (bandwidth), latency variations (jitter), and propagation latency.
Our research investigates the problem of translating the end-to-end delay requirements of the flows to a specific resource allocation at the different routers in the DiffServ model. The required end-to-end QoS level is usually specified in a Service Level Agreement (SLA). The problem is thus to find a way that maximizes the number of flows for which the SLA is satisfied in congested conditions. In particular we concentrate on delay, an important QoS parameter. The work is composed of both studying local algorithms that guarantee a constant ratio between local waiting times of packets belonging to different service classes, and studying global probing techniques that allow classifying the packets into the different service classes according to their SLA defined delay.
Our results indicate that when compared to a single FIFO queue in all routers (today’s “best-effort” situation), any local algorithm with or without probing, dramatically increases the number of satisfied packets. We also show that while a simple local resource allocation algorithm can provide a fairly good ratio between the waiting time of packets in the different classes, when considering global maximization of SLA satisfied packets, such an algorithm is not really needed.
Finally, we show that a global probing mechanism with an adaptive classification of packets to service classes, can significantly increase the number of packets for which the SLA is satisfied.