Time-Driven Priority Router Implementation: Analysis and Experiments

摘要

Low complexity solutions to provide deterministic quality over packet switched networks while achieving high resource utilization have been an open research issue for many years. Service differentiation combined with resource overprovisioning has been considered an acceptable compromise and widely deployed given that the amount of traffic requiring quality guarantees has been limited. This approach is not viable, though, as new bandwidth hungry applications, such as video on demand, telepresence, and virtual reality, populate networks invalidating the rationale that made it acceptable so far. Time-driven priority represents a potentially interesting solution. However, the fact that the network operation is based on a time reference shared by all nodes raises concerns on the complexity of the nodes, from the point of view of both their hardware and software architecture. This work analyzes the implications that the timing requirements of time-driven priority have on network nodes and shows how proper operation can be ensured even when system components introduce timing uncertainties. Experimental results on a time-driven priority router implementation based on a personal computer both validate the analysis and demonstrate the feasibility of the technology even on an architecture that is not designed for operating under timing constraints.