I would love some additional feedback on FEC. It appears to me that when evaluating FEC with test tools, generating random packet loss as a percentage of packets, FEC works as advertised. But in real life, packet loss often occurs in bunches due to WIFI handovers, route flaps, and temporal network congestion.
In a previous post, there was a wise comment about how FEC can increase latency, and its a trade off between how much FEC is applied, verses the increased bandwidth and latency. I'm a believer that dynamic FEC may never be responsive enough, or be able to anticipate events that cause packet loss in real life.
I think the ideal use case for FEC is microwave circuits. These tend to really simulate random packet loss that is consistent during a rain storm. I doubt there are any other uses cases where dynamic FEC will actually improve things much at all.
No comments:
Post a Comment