Topology matters in network coding

Seeking to understand the potential of network coding in future internet endeavors, we consider the vital role of network topology with respect to the potential benefits of Random Linear Network Coding (RLNC). First, we propose a set of metrics that capture the essential trade-offs between throughput, confidentiality and decoding delay. Using large network simulation, we are able to evaluate the behavior of RLNC for various topological classes based on random graphs. Our results show significant differences between local dissemination of information (typical of wireless networks) and long range connectivity (typical of peer-to-peer environments). We believe that our results can help pave the way for the creation of better overlay topologies for RLNC protocols in future internet applications.