Towards Fast Flow Convergence in Cognitive Radio Cellular Networks

Cognitive radio (CR) is an enabling technology that allows opportunistic use of under-utilized licensed spectrum allocated to primary users (PUs).However, the frequent channel sensing and switching interferes with the transport layer functions, leading to slow flow convergence during active transmissions by the CR. In this paper, we propose TCP C², a method that greatly improves the flow responsiveness to abrupt variation of underlying layer spectrum availability in cellular CR architectures. The key idea of C² is to allow the sender to estimate the current bottleneck link bandwidth and network load by observing variance in the throughput and round trip time. Following this, fast congestion window scaling allows the flow to converge quickly to the optimal sending rate. Analytic derivations and packet-based simulation results show the increased resiliency of our approach over classical end-to-end TCP protocols in the presence of intermittent spectrum sensing and disruptions caused by PU arrival. Additionally, we show that C² enforces fairness among flows, and also coexists well with classical TCP flavors.