Node delay assignment strategies to support end-to-end delay requirements in heterogeneous networks

In a complex, heterogeneous network environment, such as the Internet, packets traversing different networks may be subjected to different treatments and may face different traffic loads across the routing path. This paper addresses the key issue of how to assign delay budgets to each network node along the routing path so that the end-to-end delay requirements of the supported applications are met. First, we describe a methodology to compute for a given flow a set of feasible per-node delays for the class of delay-based servers. We then formalize an optimal per-node delay assignment problem which takes into consideration the workload across the routing path. The solution, for homogeneous and heterogeneous networks, is provided. The resulting solution is optimal, but its implementation overhead is relatively high. To overcome this shortcoming, we propose two heuristics, EPH() and LBH(), to approximate the optimal strategy. EPH() uses the equi-partition concept to compute initial delay values and adjust these delay values to meet the end-to-end delay requirements. LBH() uses a relaxation factor to distribute the load proportionally across all nodes on the routing path. A simulation-based comparative analysis shows that the heuristics perform closely to the optimal schemes.