CL-ADSP: Cross-Layer Adaptive Data Scheduling Policy in Mobile Ad-hoc Networks

摘要

Multi-Path Transmission Control Protocol (MPTCP), leading transport layer protocol, radically utilizes multi-path communication technology. MPTCP's vital objective is to exploit the multi-interface capability available at the end-host. Nevertheless, the protocol majorly suffers from lower channel utilization because of its abrupt congestion window (cwnd) growth policy hence, experiences soaring throughput degradation. Further, multiple available paths for transmission severely degrades MPTCP's performance as higher re-ordering is required at the receiver end because of dissimilar wireless path characteristics (e.g., bandwidth, loss rate (either due to congestion or unreliability of wireless channel) and delay). Thereby, it leads to the serious rise in both the issues of buffer-blocking and un-necessary retransmissions. Although, a few solutions have been suggested that effectively schedule and balance the load over available qualified paths. But, most of these solutions were not able to handle the redundant cwnd growth adaptations. This paper suggests applying a Cross-Layer based Adaptive Data Scheduling Policy (CL-ADSP) to MPTCP, which takes care of dissimilar path characteristics and unreasonable cwnd growth adaptations effectively. CL-ADSP uses a delay-variation-based adaptive fast retransmission policy. Moreover, CL-ADSP dynamically exploits the average MAC-layer retries as a parameter which appropriately shows the early sign of network overload. The simulation results exemplify the effectiveness of CL-ADSP concerning better throughput, improved file transfer time with lesser End-To-End (ETE) delay and number of timeout packets. In fact, when compared with MPTCP, CL-ADSP offers 24% better throughput with 36% lesser timeout packets, while, 10% improved packet delivery ratio with 2.25% lesser average ETE delay results. CL-ADSP also offers 23% improved file transfer time and 24% lesser normalized routing load over MPTCP. (C) 2019 Elsevier B.V. All rights reserved.