[ndnrtc-arc] Data transfer encounters failure in long-delay environment
Problem description (per Yoneda-san):
Consumer stops receiving data over long-delay links after some time.
Consumer slowly "drifts away" from producer in issuing challenge interests. In other words, there is accumulating error in frame numbers which are used by consumer for issuing interests for a higher bitrate thread, it manifests itself in producer receiving more “old” interests (i.e. interests for the data that has already been outdated).
Here are some analysis findings:
- Interest-Data exchange over time: data receipt declines (char 1.png attached)
- same exchange, but per thread: now we see that “mid” thread suffers (this is the thread used by ARC for challenging data) (chart 2.png attached)
- amount of new/old incoming interests on producer side (overall and per thread): one can see that mid thread received more old interests over time, while “low” thread didn’t get at all (that’s expected). In general, there should be no “old” interests at all (chart 3.png attached)
Updated by Peter Gusev almost 5 years ago
- File consumer-yoneda-main_camera-delay-test4.tar.gz consumer-yoneda-main_camera-delay-test4.tar.gz added
This is log file in case of 400ms(RTT) environment.
In this case, NDNRTC cannot shit fetching state, and then ARC is never called.
Each link (L1 , L2) is 100Mbps, 200ms(both way)