Comparing Rely and Reed Solomon FEC/ECC
One question that we often get is how do different ECC/FEC algorithms compare? It’s a simple question with a complicated answer - the reason is that different ECC/FEC algorithms are designed with different goals in mind. To simplify the answer we need to restrict the scenario in which we compare performance.
In this post we will compare two algorithms, namely Rely and Reed-Solomon. Rely is a sliding window based RLNC algorithm, whereas Reed-Solomon is based on a classical Vandermonde matrix construction.
Comparison
The basic comparison metric for the two algorithms is:
How much repair traffic is required to reach a target residual packet loss rate of 0.001%?
Scenario
Setup (also used in 3GPP MBMS FEC evaluation) refer to [1] for more details:
LTE MBMS Bearer bitrates: 1.0656 Mbps
Radio Link Control - SDU size: 1332 bytes
Radio Link Control - SDU frequency: 10ms
Constant bitrate channel (100 packets per second)
Latency budget 240 ms (or 24 packets)
We use a random uniform loss model which is different from the one used in [1], and compare the two algorithms using the following simulation setup:
The traffic generator produces 600,000 data packets which are sent via the encoder over the lossy channel and into the decoder. The residual packet loss after decoding is then recorded. In order to find the required repair rate we sweep over multiple configurations (low to high) until we find a configuration that can meet the target residual packet loss rate.
Results
The following figure show the results:
There a few key takeaways:
Notice that even with a very limited latency budget (only 24 packets). Rely based on sliding window RLNC significantly outperforms the block based Reed-Solomon code.
For losses over 10% Rely reduces the required repair traffic by more than 13%.
Conclusion
These results illustrate some of the benefits of Rely with respect to other coding schemes. To learn more about Rely head over below and as always if you have questions about this article or FEC/ECC in general feel free to reach out to us on contact@steinwurf.com.
References
[1] ETSI, “Evaluation of mbms fec enhancements (final report),” Dec. 2015, 3GPP TR 26.947 version 13.0.0 Release 13.