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

Schütz, Bertram; Aschenbruck, Nils; Bussmann, Stephan; Döring, Hanna; Juhnke-Kubitzke, Martina

doi:10.1109/lcnsymposium50271.2020.9363259

Cited by 2 publications

(1 citation statement)

References 14 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…Since popular link emulation tools like netem do not take the length of a packet into account and in order to keep the computation time reasonably low, we base our numerical analysis on a Benoulli channel model. A more complex evaluation of the impact of bit errors on packets with heterogeneous lengths is conducted in [31]. To calculate the packet loss rate that remains after decoding, P LR DEC , following formula is used: with:…”

Section: Numerical Parameter Determinationmentioning

confidence: 99%

Less Is More: Choosing Fair Network Coding Parameters

Thieme

Schütz

Aschenbruck

2021

2021 IEEE 46th Conference on Local Computer Networks (LCN)

Self Cite

View full text Add to dashboard Cite

TCP still is the prevailing transport protocol on the Internet. It does, however, suffer from severe performance issues on links with non-congestion losses. Instead of relying on the RTT-sensitive ARQ mechanism of TCP, network coding (NC) uses linear combinations of packets to correct forward erasures. Several approaches were developed to incorporate NC into TCP. While they are able to improve the offered service, their lack of deployability and fairness in modern environments is not desirable. To mitigate this, we propose TCPyNC, a Python-based intra-session NC user space shim layer for TCP. Not only are its gains equally high, but it is also more flexible regarding the use of TCP options and parameters. We use a numerical analysis and a testbed evaluation to determine parameters that enable a fair use of TCP in conjunction with Forward Erasure Correction (FEC). In order to reach this goal, we found the maxim less is more to be a sensible orientation. For scenarios with 10% random loss, instead of choosing a high redundancy rate and a large generation size in order to gain the best loss reduction, choosing redundancy rates that leave between 1% and 6% loss for TCP's ARQ to correct and generation sizes below 32 when using default TCP parameters, yields fair results.

show abstract

Section: Numerical Parameter Determinationmentioning

confidence: 99%