FAST CLOSED FORM APPROXIMATION FOR DYNAMIC NETWORK RESOURCE ALLOCATION
J. Joutsensalo, Isto Kannisto, Jarmo Siltanen, Timo Hämäläinen · 2006
We consider dynamic delay guarantee and bandwidth allocation in communications networks. Our scenario includes linear pricing scheme for both Quality of Service parameters. The goal is (i) to maximize the revenue and (ii) guarantee fair resource allocation for connections. On the contrary to the traditional Lagrangian approach, we approach the problem by modified one, where the sum of the weights of the scheduler acts as the penalty term. This modified approach yields closed form approximate algorithm for updating the scheduler weights, being very fast and real-time implementable. We compare the algorithm with the brute-force method, which optimizes weights in the large grid- optimal brute-force method has exponential complexity. The revenue obtained by the closed form method is about 99.9 % of the optimal, computationally expensive approach, thus being tempting both from the point of view of the service provider and the customers. NS-2 simulator is used in the experiments.