Max-min Utility Fairness in Link Aggregated Systems

Abstract

We consider utility max-min fairness of application traffic of a next generation Internet (NGI) service model contending for network bandwidth. The physical infrastructure that offers the bandwidth is a link-aggregated system where the links can be homogeneous or heterogeneous. Traffic can broadly be categorized into a few distinct classes and the containing flows in each class utilize the available bandwidth in the same way. The basis of this categorization is the realization that different applications derive different utility even when alloted the same bandwidth. The goal is to allocate bandwidth so as to equalize the utilities to each class is in a maximal and fair way. Once the required rates are computed these are utilized by the scheduler for scheduling packets over the links.