CellReplay: Towards accurate record-and-replay for cellular networks
summary
CellReplay records cellular behavior with two simultaneous workloads — light packet trains and a saturator — calibrates parameters, and replays by interpolating packet delivery schedules into a Mahimahi-compatible emulator using two phones and an automated calibration phase. Evaluation with web browsing, TCP downloads, and video ABR across mobility shows it reduces emulation error from 18.77% to 5.68%, corrects Mahimahi’s bandwidth overestimation, and reproduces live-network page loads and bitrates far more faithfully.
AI-written summary · about
chapters (19)
- 0:00CellReplay introduction
- 1:19Live 5G testing limits
- 2:43Mahimahi record-and-replay
- 4:06Mahimahi vs live networks
- 4:48Emulation inaccuracies causes
- 5:28Fixed delay problem
- 6:09Bandwidth workload dependency
- 6:51Throughput measurement
- 7:53Packet delivery opportunities
- 8:59Multi-workload recording
- 9:42Heavy vs light workloads
- 10:24How packet delivery opportunity interpolation works
- 11:07CellReplay framework
- 12:11CellReplay vs Mahimahi
- 13:16Emulation setup method
- 13:36Web and file accuracy
- 14:18ABR algorithm results
- 15:22Non-stationary evaluations
- 15:44Conclusion and future
related talks
19mSolve your Mobile Bugs with Session Replay
26mDETER: Deterministic TCP Replay for Performance Diagnosis
15mInside Mobile Session Replay: Smarter Debugging
15mREAL: Emulating Control Plane at Simulator’s Cost
17mAUGUR: Practical Mobile Multipath Transport Service for Low Tail Latency in Real-Time...
19mDChannel: Accelerating Mobile Applications With Parallel High-bandwidth...
26mVisualizing the Unseen to Fix Errors Faster with Sentry’s Session Replay Live
16mQCON: Seamless QoE-Aware 5G Streaming via Multi-Connectivity