TY - JOUR
T1 - Fulcrum
T2 - Flexible network coding for heterogeneous devices
AU - Lucani, Daniel E.
AU - Pedersen, Morten V.
AU - Ruano, Diego
AU - Sorensen, Chres W.
AU - Fitzek, Frank H.P.
AU - Heide, Janus
AU - Geil, Olav
AU - Nguyen, Vu
AU - Reisslein, Martin
N1 - Publisher Copyright:
© 2018 IEEE.
PY - 2018
Y1 - 2018
N2 - We introduce Fulcrum, a network coding framework that achieves three seemingly conflicting objectives: 1) to reduce the coding coefficient overhead down to nearly n bits per packet in a generation of n packets; 2) to conduct the network coding using only Galois field GF(2) operations at intermediate nodes if necessary, dramatically reducing computing complexity in the network; and 3) to deliver an end-to-end performance that is close to that of a high-field network coding system for high-end receivers, while simultaneously catering to low-end receivers that decode in GF(2). As a consequence of 1) and 3), Fulcrum has a unique trait missing so far in the network coding literature: Providing the network with the flexibility to distribute computational complexity over different devices depending on their current load, network conditions, or energy constraints. At the core of our framework lies the idea of precoding at the sources using an expansion field GF(2h),h > 1, to increase the number of dimensions seen by the network. Fulcrum can use any high-field linear code for precoding, e.g., Reed-Solomon or Random Linear Network Coding (RLNC). Our analysis shows that the number of additional dimensions created during precoding controls the trade-off between delay, overhead, and computing complexity. Our implementation and measurements show that Fulcrum achieves similar decoding probabilities as high field RLNC but with encoders and decoders that are an order of magnitude faster.
AB - We introduce Fulcrum, a network coding framework that achieves three seemingly conflicting objectives: 1) to reduce the coding coefficient overhead down to nearly n bits per packet in a generation of n packets; 2) to conduct the network coding using only Galois field GF(2) operations at intermediate nodes if necessary, dramatically reducing computing complexity in the network; and 3) to deliver an end-to-end performance that is close to that of a high-field network coding system for high-end receivers, while simultaneously catering to low-end receivers that decode in GF(2). As a consequence of 1) and 3), Fulcrum has a unique trait missing so far in the network coding literature: Providing the network with the flexibility to distribute computational complexity over different devices depending on their current load, network conditions, or energy constraints. At the core of our framework lies the idea of precoding at the sources using an expansion field GF(2h),h > 1, to increase the number of dimensions seen by the network. Fulcrum can use any high-field linear code for precoding, e.g., Reed-Solomon or Random Linear Network Coding (RLNC). Our analysis shows that the number of additional dimensions created during precoding controls the trade-off between delay, overhead, and computing complexity. Our implementation and measurements show that Fulcrum achieves similar decoding probabilities as high field RLNC but with encoders and decoders that are an order of magnitude faster.
KW - Decoding probability
KW - random linear network coding (RLNC)
KW - resource-constrained devices
KW - throughput
UR - http://www.scopus.com/inward/record.url?scp=85057772210&partnerID=8YFLogxK
UR - http://www.scopus.com/inward/citedby.url?scp=85057772210&partnerID=8YFLogxK
U2 - 10.1109/ACCESS.2018.2884408
DO - 10.1109/ACCESS.2018.2884408
M3 - Article
AN - SCOPUS:85057772210
SN - 2169-3536
VL - 6
SP - 77890
EP - 77910
JO - IEEE Access
JF - IEEE Access
M1 - 8554264
ER -