Low-latency Sliding-window Recoding

Abstract

In wireless multi-hop and multi-path computer networks, such as the ones that potentially emerge with Internet of Things (IoT) deployments, channel coding has become a tried and tested solution to handle packet losses. Random Linear Network Coding (RLNC) is one of such block coding techniques, which also provides the unique feature of recoding already coded symbols on intermediate nodes, thereby increasing the general throughput of the network. However, the computation overhead of decoding on the receiver device counteracts the much desired low-latency requirements of various fifth generation (5G) use cases. Sliding-window RLNC is a technique, which can minimize such overheads. Nonetheless, recoding over a sliding window destroys the low-latency decodable properties of the scheme. In this paper we present a low-latency recoding technique that balances packet forwarding and recoding on the intermediate nodes in way that potentially preserves the sliding window characteristics. Our simulation results show that the performance of the proposed recoding algorithm achieves a two factor improvement in latency and losses, as well as a more than three times gain in throughput.