Dynamic Resource Allocation in Core Routers of a Diffserv Network

Abstract

The Differentiated Services (DiffServ) architecture is receiving wide attention as a framework for providing different levels of service according to a Service Level Agreement (SLA) profile. The edge routers in a DiffServ network mark/shape/police flows based on their SLAs, and the core routers offer packets different treatments using the marks they carry. Core routers handle aggregates of flows instead of individual flows, which is known to considerably reduce the complexity of DiffServ. Tuning core routers is clearly an important issue to satisfy the needs of traffic marked at the edges. This tuning is actually characterized by extensive manual work, based on a trial-and-error process, which is often ineffective, time-consuming and costly to network managers. In this paper, we propose a dynamic, self-tuning mechanism for allocating resources in core routers among Diffserv services. Our mechanism is easy to implement, and does not require any particular signaling. It ensures that SLAs are respected, and allows at the same time an efficient utilization of network resources. We validate the performance of our mechanism by a campaign of experiments on a real network testbed.