A Practical Congestion Control Scheme for Named Data Networking

Traditional congestion control mechanisms are designed for end-to-end connections and do not fit the Named Data Networking (NDN) architecture, in which content can be retrieved from multiple sources and through multiple paths. To fully exploit the NDN architecture, a congestion control scheme must consider the effects of in-network caching, multipath forwarding, and multicast data delivery. Moreover, the solution must not assume known link bandwidths or Data packet sizes, as these assumptions may not hold for overlay links, wireless links, or applications with varying Data packet sizes. In this paper we propose PCON: a practical congestion control scheme to address the above issues. PCON detects congestion based on the CoDel AQM (by measuring packet queuing time), then signals it towards consumers by explicitly marking certain packets, so that downstream routers can divert traffic to alternative paths and consumers can reduce their Interest sending rates. Our simulations show that PCON's forwarding adaptation reaches a higher total throughput than existing work while maintaining similar RTT fairness. Moreover, PCON can adapt to the changing capacity of IP tunnels and wireless links, conditions that other hop-by-hop schemes do not consider.