Adaptive low-priority congestion control for high bandwidth-delay product and wireless networks

Abstract

The low-priority service is an exciting and attractive choice for networking applications (e.g. automatic update, backup, peer-to-peer file share) which create traffic that is considered less urgent than that of others and become a renewed interest at the Internet Engineering Task Force (IETF). A low-priority protocol, which provides the low-priority service, can exploit the residual bandwidth of the bottleneck link and achieve the high throughput, low-latency data delivery in traditional networks. However, due to the conservative and inappropriate congestion control mechanisms, the existing low-priority protocols (e.g. LEDBAT) cannot effectively utilize the residual bandwidth of the bottleneck link in high bandwidth-delay product (HBDP) and wireless networks. In this paper, we propose an adaptive Congestion level-based Low-Priority congestion Control (CLPC) protocol to improve the efficiency of low-priority protocols and maintain the low-priority features. Specifically, the CLPC sender adopts an one-way path delay to estimate the congestion level and adjust the aggressiveness of congestion control mechanisms. Different from other low-priority protocols, the CLPC protocol is more aggressive when the bottleneck link of HBDP and wireless networks has residual bandwidth. This makes a faster convergence of link utilization in HBDP networks. Combining the random loss detection, CLPC can achieve the high throughput in wireless networks. The extensive simulations in NS-2 show that CLPC can improve the transmission performance significantly as compared to other low-priority protocols in HBDP and wireless networks. Furthermore, we implement the CLPC protocol in the Linux kernel (3.13) and setup a testbed to measure it. The results also indicate the feasibility and effectiveness of CLPC.