The Impact of Bit Errors on Intra-session Network Coding with Heterogeneous Packet Lengths

While intra-session network coding has proven to be a promising packet-level forward error correction scheme, recent studies have shown that it can invoke increased packet sizes and excessive overhead. This problem is especially significant for traffic with heterogeneous packet lengths. Yet, most conducted studies emulate packet loss using length-independent random loss models, ignoring the overhead. In reality, larger packets have a higher probability to get corrupted, e.g., by failing a series of link layer cycling redundancy checks. To analyze the impact of this problem, the here presented study instead applies a bit error model to the encoded traffic, including the coding overhead. We will present formal equations to calculate the impact in terms of frame error rate and packet loss rate before and after decoding. Finally, a trace-based simulation highlights the implications for relevant real-world applications, namely QUIC web browsing, UFTP file transfer, Skype VoIP, and FFmpeg video streaming.