Abstract

We are designing a simple measurement-based admission control algorithm that allows the edge router to track essential traffic rate parameters by which comprehensive QoS-based admission control can be conducted.  The main advantage of our approach is that it allows the network to adapt to the traffic rate characteristics including particularly the degree of self-similarity.  While applying the resource allocation algorithm developed in the previous year to guarantee average throughput and source queue delay on a per-flow basis, scheduling delay at the edge router and packet loss ratio must be managed by a comprehensive scheduling algorithm at the edge router.  We have designed such an algorithm that controls the outgoing (to the core) traffic in accordance with the service level agreement, enforces differentiated queuing delay guarantees at the edge router, and manages differentiated packet loss ratio through A measurement-based admission control algorithm.  

The scheduling algorithm features configurability, scalability, flexibility and analytical tractability.  Basically, the algorithm captures the essence of fair queuing and random early detection on one hand while delivering differentiated QoS on the other.  We expect to derive close-form analytical expressions for bandwidth efficiency, for mean scheduling delay and for loss ratio for every service class.  These results will be useful for designing QoS-driven admission control policies and for constructing convenient spreadsheet tools for capacity analysis and for analyzing performance tradeoffs.

Students

• Jeorg Furrer

• Yishen Sun