Massive parallelization technique for random linear network coding

Abstract

Random linear network coding (RLNC) has gain popularity as a useful performance-enhancing tool for communications networks. In this paper, we propose a RLNC parallel implementation technique for General Purpose Graphical Processing Units (GPGPUs.) Recently, GPGPU technology has paved the way for parallelizing RLNC; however, current state-of-the-art parallelization techniques for RLNC are unable to fully utilize GPGPU technology in many occasions. Addressing this problem, we propose a new RLNC parallelization technique that can fully exploit GPGPU architectures. Our parallel method shows over 4 times higher throughput compared to existing state-of-the-art parallel RLNC decoding schemes for GPGPU and 20 times higher throughput over the state-of-the-art serial RLNC decoders.