Heterogeneous quality of service in DiffServ - static and dynamic multicast
Anantha Padmanaban, Sai Sudhir (2004) Heterogeneous quality of service in DiffServ - static and dynamic multicast. Masters thesis, Iowa State University.
Full text available as:
Multicasting in DiffServ networks is a challenging problem due to inherent architectural conflicts: stateful core (multicast) vs. stateless core (DiffServ). This thesis addresses the problem of heterogeneous QoS multicasting in DiffServ assuming an edge-based multicast model and makes the following two contributions. First, we formulate the problem of static multicast in heterogenous QoS in DiffServ wherein we provide a rigorous formulation of the problem as a network optimization problem and propose two algorithms, namely, Merging KMB Trees (MKT) and Highest Level In First (HLIF) to solve it. While MKT is an approximation algorithm for which we prove a worst-case bound, HLIF is a heuristic. We also evaluate the performance of the algorithms through simulation studies using a new metric, "Tree QoS Cost" by comparing them with a well known multicast tree construction algorithm, minimum-cost path heuristic. We also study the resulting "good neighbor effect" of the algorithms. Our studies show that the tree QoS cost of HLIF is better than that of MPH and MKT, and the good neighbor instances of HLIF and MKT are less than that of MPH. We also show that the empirical performance of HLIF and MKT are signi¯cantly better than the theoretical worst-case bound on MKT. The second contribution of this thesis is to solve the problem of dynamic multicast in heteregeneous QoS in DiffServ wherein we formally state the problem and propose an optimal algorithm for it which does not incur any service disruption due to group dynamics. We evaluate the performance of our algorithm with HLIF and a source-based shortest path algorithm using "Tree QoS Cost" and "Number of Service Disruptions" as metrics. Our studies show that the proposed algorithm achieves good performance in terms of Tree QoS Cost, time taken for member join/leave, and number of service disruptions with acceptable storage overhead.
Archive Staff Only: edit this record