Abstract

In this work we study the limitations of on-line schemes used to deal with the interplay of latency and utilization parameters in bandwidth allocation for tasks with varying and unpredictable bandwidth demands. We provide tight latency lower bounds to any randomized adaptive bandwidth allocation scheme for a single worst-case input stream of packets (session). We achieve this by employing the min-max principle of Yao and by characterizing the optimal (with respect to average queue size) deterministic scheme, which is subjected to a suitably chosen (\worst case") random input. This submission should be considered for both regular and brief announcement tracks. This work was partially supported by the EU Projects ESPRIT LTR ALCOM- IT, IST-FET-OPEN ALCOM-FT , IMPROVING RTN ARACNE and the Greek GSRT Project PENED-ALKAD. 1 1 Introduction Current (and future) communication networks are inviting an ination in their bandwidth demand, due to users request for quality of service (...

Citations