From Source to Solution: Tackling Packet Losses in Large-scale Cloud Gaming...
summary
Production measurements on Tencent START (≈200,000 hours) show 66.5% of losses arise from network bias against UDP and congestion; LADR implements a TCP-first multi-connection proxy, combines delay-based Pega with a loss-based bitrate controller (LBBC), and replaces receiver NACK with sender-side SACK/RACK-TLP plus opportunistic FEC and reactive rate limiting. Deployed for a year, LADR reports 0.049% packet loss (−59.8% vs. prior) and delivers 99.87% of frames within 100 ms.
AI-written summary · about
chapters (13)
- 0:00Cloud gaming packet losses
- 1:01Loss detection and recovery
- 2:49Production stack and measurements
- 4:34WebRTC assumptions revisited
- 5:39Transport protocol choice and TCP
- 6:40Congestion control for shallow buffers
- 7:22Loss detection timing and SACK
- 8:25FEC limitations and recovery
- 9:06Designing LADAR with loss sources
- 10:12Multi-connection transport with proxy
- 11:14Production results and tail reduction
- 13:45Summary and research context
- 15:32Acknowledgments
related talks
19mPudica: Toward Near-Zero Queuing Delay in Congestion Control for Cloud Gaming
18mRevisiting Congestion Control for Lossless Ethernet
17mRe architecting Congestion Management in Lossless Ethernet
20mETC: An Elastic Transmission Control Using End-to-End Available Bandwidth...
22m007: Democratically Finding the Cause of Packet Drops
21mNetworking for Distributed Cloud & Edge Computing
29mNetwork Fault Finding System: Packet Loss Triangulation
15mEnabling High-Quality Real-Time Communications with Adaptive Frame-Rate