Tambur: Efficient loss recovery for videoconferencing via streaming codes
summary
Tambur applies streaming codes tuned to videoconferencing, evaluated on Microsoft Teams traces and in emulated end-to-end calls after being implemented in C++ and integrated with a conferencing app. The system uses per-frame parity and a lightweight classifier to pick redundancy, recovering bursts within a 3-frame deadline while still fixing single-packet losses immediately. Compared to the baseline, Tambur yields 26% fewer decoding failures and 35% less redundancy in simulation and cuts freeze frequency and duration by about 26% and 29% in end-to-end tests.
AI-written summary · about
chapters (9)
- 0:00Start
- 2:01Conventional loss-recovery is ill-suited to VC
- 3:37RS code within each frame wastes parity
- 7:33Streaming codes: bandwidth-efficient loss recovery
- 8:23Streaming codes: challenges
- 9:16Analysis of traces from Teams video calls
- 16:04Online evaluation methodology
- 17:56Tambur mitigates freeze frequency
- 18:41New interdisciplinary loss recovery VC
related talks
17mHairpin: Rethinking Packet Loss Recovery in Edge-based Interactive Video Streaming
28mSalsify: Low-Latency Network Video through Tighter Integration
20mReconstructing proprietary video streaming algorithms
21mLearning in situ: a randomized experiment in video streaming
16mFrom Source to Solution: Tackling Packet Losses in Large-scale Cloud Gaming...
20mETC: An Elastic Transmission Control Using End-to-End Available Bandwidth...
14mEndangered Privacy: Large-Scale Monitoring of Video Streaming Services
17mBackdraft: a Lossless Virtual Switch that Prevents the Slow Receiver Problem