Congestion Control using FEC for Conversational Multimedia Communication

Presented by: Marcin Nagy
at ACM International Conference on Multimedia Systems, 2014
Singapore

Co-authors:
- Varun Singh, Joerg Ott (Aalto University)
- Lars Eggert (NetApp)

Transcript

1. 1 Congestion Control using FEC for Conversational Multimedia Communication Marcin Nagy, Varun Singh, Jörg Ott Aalto University {firstname.lastname@aalto.fi} Lars Eggert NetApp, Germany lars@netapp.com
2. 2 Problem How can we use Forward Error Correction (FEC) for error resilience… … and rate control simultaneously
3. 3 Outline •  Conversational multimedia introduction •  Current congestion control approaches •  Our approach: using Forward Error Correction
4. 4 Outline •  Conversational multimedia introduction •  Current congestion control approaches •  Our approach: using Forward Error Correction
5. 5 Conversational communication Network Video, FEC, RETX à ß RTCP Feedback •  Heart of emerging RTC-enabled web applications •  Strict delay on end-to-end packet delivery - 400ms (3GPP standard) •  Small playout buffer - of magnitude smaller than for streaming
6. 6 RTP and RTCP Sender Receiver RTP media stream (encoded media, FEC, repair) RTCP Sender Reports (SRs) •  Timing, synchronization •  Sending rate, packet count RTCP Receiver Reports (RRs) •  Rough statistics •  Congestion cues RTCP XRs: •  Detailed Statistics •  Dejittering, sync, playout •  Monitoring + reporting •  Event notifications •  Local error concealment Short-term adaptation •  Error-resilience (NACK, PLI) •  Congestion control •  Adaptive source coding Long-term adaptation •  Codec choice •  Packetization size •  FEC, interleaving
7. 7 Error resilience mechanisms NACK (RTX) RPS SSA UEP 3G Link Packet Loss End-to-End Delay •  NACK •  Reference Picture Selection •  Slice Size Adaptation •  Uneven Level Protection (FEC)
8. 8 Forward Error Correction (FEC) •  Redundant stream on top of media stream •  Used to recover lost packets •  Various schemes available: XOR-based parity, …
9. 9 Outline •  Conversational multimedia introduction •  Current congestion control approaches •  Our approach: using Forward Error Correction
10. 10 Current congestion control solutions •  TFRC •  RAP •  RRTCC •  C-NADU •  NADA •  DFLOW •  … •  MSFTP (uses FEC but for streaming) •  SVC Congestion control with error resilience decoupled
11. 11 Outline •  Conversational multimedia introduction •  Current congestion control approaches •  Our approach: using Forward Error Correction
12. 12 Can we use FEC for rate control? •  Primarily the error protection method •  Rate control aims at optimal network path utilization –  Mitigates risk of congestion losses •  Rate control requires probing for capacity •  Idea: 1.  Use FEC to probe for available bandwidth 2.  Additional stream protection è more aggressive probing
13. 13 FEC rate control concept RTCP Interval time Throughput media rate FEC STAY state! PROBE state! STAY state! STAY state! UP state! UP state! DOWN state! PROBE state! (A) Adding FEC (B) Swapping FEC with media (C) Partially swapping FEC with media Available Capacity
14. 14 FEC rate control state machine STAY UP DOWN PROBE Good conditions, probe network No more bandwidth available Stable conditions Stable conditions Unstable conditions Unstable conditions Unstable conditions
15. 15 What is the right FEC rate? Intuition: –  High FEC rate •  Higher error protection •  Quick sending rate ramp-up •  Network over-utilization risk –  Low FEC rate •  Lower error protection •  Slow sending rate ramp-up •  Minimizes capacity overshooting Concept: 1.  High FEC rate when operating far from capacity limit 2.  Low FEC rate when approaching capacity limit
16. 16 FEC Based Rate Adaptation Algorithm •  Delay-based algorithm •  Congestion cues: –  RTCP XR (losses and discards) –  One-way delay (OWD) [Ngamwongwattana et. al. 2010] •  XOR-based parity FEC scheme –  2-14 media frames protected by one FEC frame •  Special congestion handling procedures: –  Undershooting –  Bounce-back
17. 17 Evaluation 0 50 100 150 200 250 300 350 400 0 100 200 300 400 500 600 700 800 900 Throughput [kbps] Time [s] •  Ns-2 simulation •  Variable link capacity •  50ms one-way-delay •  One flow on the link
18. 18 Evaluation •  Dummynet emulator •  1Mbit/s link capacity •  50ms one-way-delay •  Two RTP flows on the link •  GOP size = 12 0 200 400 400 200 0 0 50 100 150 200 250 300 350 400 0 50 100 50 0 Media Throughput [kbps] FEC Rate [kbps] Time [s] Media 1 Media 2 FEC_Rate 1 FEC_Rate 2
19. 19 Next steps? •  Generalize concept to any congestion control –  RRTCC with FEC
20. 20 Summary •  A new rate control approach based on FEC •  Generic idea that can be incorporated to other solutions –  e.g., RRTCC •  Efficient new congestion control algorithm •  Increased GOP size provides energy savings •  Ready to deploy
21. 21 Outline •  Conversational multimedia introduction •  Current congestion control approaches •  Our approach: using Forward Error Correction •  Adaptive Multimedia System (AMuSys)
22. 22 AMuSys •  Multimedia congestion control never stand-alone •  Tight-cooperation between components required •  Adaptive Multimedia System (AMuSys) –  3 sub-systems •  application, codes, networking components –  Control loops [Singh et. al. 2012] –  Easy to use interface for app developers
23. 23 AMuSys Expectations: -camera -codec -frame rate Obtain: - reception stats - received content - RTP/RTCP - Capacity est. - set parameters - change bitrate - get parameters - get bitrate