Network coding simplifies routing decisions, improves
throughput, and increases tolerance against packet loss. A
fundamental limitation, however, is delay: decoding requires as
many independent linear combinations as data blocks. Prioritized
network coding reduces this delay problem by introducing a
hierarchy of prioritization layers. What remains is the problem
of choosing a layer to approach two often-contradicting goals:
reduce delay until prioritized layers can be decoded and keep
the total number of transmissions low. In this paper, we propose
an algorithm for this problem that – based on limited feedback –
primarily minimizes per-layer delay but identifies opportunities
to reduce the required transmissions when per-layer delay is
unaffected. Our evaluation shows that our algorithm improves
per-layer delay compared to hierarchical network coding and is
close to the theoretical optimum number of total transmissions.
