Analysis and Evaluation of Fully TCP-Compatible Backpressure-Driven Traffic Engineering

Throughput optimality is a core objective for the operation of modern networks. The backpressure routing and scheduling scheme is well-known for being analytically-optimal in terms of throughput maximization. Its operation is driven by packet queue build-ups at the network nodes, and constant status updates among neighbors in order to deduce the best next packet forwarding step. However, this operating principle is also incompatible to TCP, which uses congestion control to avoid queue build-ups. These factors have so far limited the use of backpressure at the edge only. The present work presents a novel combination of TCP and backpressure at the network layer, overcoming any compatibility issues and allowing for backpressure to be active at all network nodes without exceptions. The key idea is to apply the Lyapunov framework cumulatively, over a large set of router hardware clock ticks. This approach allows for the throughput-optimal backpressure decisions to be expressed via regular traffic engineering routing rules, instead of explicit per-packet forwarding directives. Simulation results validate the analytical findings, and demonstrate significant throughput gains over state-of-the-art related approaches. Proof of concept emulations showcase the simplicity in the implementation of the proposed scheme.