Adaptive Flow Assignment and Packet Scheduling for Delay-Constrained Traffic Over Heterogeneous Wireless Networks

Abstract

Flow assignment and packet scheduling are two processes of deploying limited wireless resources to support high-data-rate transmissions. Optimizing these processes is critical in improving the traffic performance and network utilization. In this paper, we investigate the flow assignment and packet scheduling for multihomed communication of delay-constrained traffic over heterogeneous wireless networks. With regard to the inherent channel diversity and unreliability, large end-to-end delay and burst packet losses pose crucial challenges to guarantee goodput performance of real-time traffic. To address these critical issues, we present an Adaptive Flow Assignment and Packet Scheduling (AFAPS) framework. First, we introduce a “horizontal water filling” algorithm, which effectively integrates the channel resources in heterogeneous wireless networks to maximize the aggregate goodput. Second, we propose an alternative path interleaving scheme, which spreads out the packets' departures over multiple communication paths within the delay constraint to mitigate burst losses. The performance of the proposed AFAPS framework is evaluated through semi-physical emulations in EXata involving real Internet traffic traces. Experimental results show that AFAPS outperforms existing multipath data distribution models, in terms of goodput and end-to-end delay.