Massive Parallelization for Random Linear Network Coding

In this paper, we propose a general-purpose graphics processing unit (GPGPU) based parallelization technique for random linear network coding (RLNC). RLNC is recognized as a useful tool for enhancing performance of networked systems, and several parallel implementation techniques have been proposed in the literature to overcome its high computation overhead. However, existing parallel methods cannot take full advantage of GPGPU technology on many occasions. Addressing this problem, we propose a new RLNC parallelization technique that can exploit GPGPU architectures in full. Our method exhibits as much as a 5x increase in throughput compared to existing parallel RLNC decoding schemes leveraging GPGPU.