Optimal speed scaling for multiclass fluid queues
Résumé
Inspired by speed-scaling techniques used in servers, we investigate the energy-delay tradeoff in bandwidth-sharing networks in which nodes can regulate their speed according to the load of the system. Assuming that the network is initially congested, we investigate the rate allocation to the classes that drains out the network with minimum total energy and delay cost. We formulate this optimal resource allocation problem as a Markov decision process which proves to be both analytically and computationally challenging. We propose to solve this stochastic problem using a deterministic fluid approximation. For the case of a linear network with two links, we provides numerical evidences in the support of the fluid model as a good approximation to the stochastic control problem. For a single link shared by an arbitrary number of classes, using Pontryagin's Maximum Principle we show that the optimal-fluid solution follows the well-known cµ rule and give an explicit expression for the optimal speed.
Domaines
Informatique [cs]
Origine : Fichiers produits par l'(les) auteur(s)