Evaluation of forward prediction scheduling in heterogeneous access networks

Abstract

With widespread deployment of often overlapping mobile and wireless coverage today, modern-day smartphones may have the capability to transmit data using multiple parallel paths to improve overall communication throughput. However, multipath data transfers have serious compatibility issues with reliable transport layer protocols which dictate the data to be received at the receiver in strict sequenced order. Failing which, multipath transfers cause out-of-order data reception leading to higher data buffering and reordering costs as well as packet retransmissions. Forward Prediction Scheduling (FPS) is a data scheduling mechanism that deals with multipath data transfers by predicting which data to transmit on each available path so that the problems associated with the out-of-order data reception are avoided. This paper presents performance evaluation of FPS scheduling mechanism on a Linux-based mobile terminal called Neo FreeRunner using a variety of multiple interfaces, in particular 3G, WiFi and Ethernet. The FPS mechanism is implemented in a Linux-kernel based version of Stream Control Transmission Protocol (SCTP).