1. Streaming Video over TCP with Receiver-based Delay Control
Shruthi Victor Maria
Amrita Chowdhury

2. Roadmap Introduction Traditional TCP Exact RDC 1 bit RDC
Simulation Results
Layered Video Streaming
Conclusion

3. INTRODUCTION
TCP , a dominant transport layer protocol in current Internet.
Good congestion control capabilities.
Unicasting video streams over TCP connections: A Real Challenge
Delay and Throughput Variation of TCP connection –a major problem for video source

4. OBJECTIVE
Receiver Based Delay Control(RDC) mechanism with TCP solves the problem..
Complexity of Congestion Control Algorithm of TCP left unaltered
Changes only the TCP Receiver Side by extending the ACK Delay

5. RDC Demonstrating Traditional TCP, Exact RDC and 1 Bit RDC
Comparison between traditional TCP and Exact RDC
Comparison between Exact RDC and 1Bit RDC

6. TRADITIONAL TCP Router includes FIFO Buffer.
Incoming packets forwarded to the buffer and drained to the outgoing link
Buffer occupancy increases when arrival rate exceeds departure rate
Packet dropped when buffer is exhausted

7. EXACT RDC Router includes FIFO Buffer and the token bucket
Keeps the occupancy of the FIFO low by using the token bucket
Token bucket used to calculate the delay for each packet and the delay appended to the outgoing packet
Receiver delays the ACK depending on the delay value

8. EXACT RDC Token buffer drain rate < outgoing link rate
Packet Delay computed as
Dlocal=Ttransmission X (Otb-Ofifo)
Ttransmission Packet transmission time
Otb-Token Bucket Level
Ofifo-FIFO Level
D=Max(Dlocal, Dincoming)
Dincoming-Incoming appended Delay

9. Properties of Exact RDC
Reduced Number of Timeout –TCP Rate =cwnd/RTT
To reduce TCP Rate- RTT is increased
Reduce Network Queueing Delay

10. 1 Bit RDC
Delay is encoded in 1 bit determined by the difference between the token buffer level and FIFO buffer level
The receiver adjusts the ACK delay based on the percent of the received packets with the CE bit set

11. 1 Bit RDC
1 bit RDC is based on multiplicative increase and additive decrease of delay
Dn+1 = 2 X Dn Dn+1> Dn
Dn+1 = (cwnd/(cwnd+1)) X Dn Dn+1<Dn

12. Exact RDC v/s 1 Bit RDC
In exact RDC, calculated delay reflects the exact congestion level and the receiver can quickly adjust.
In 1 bit RDC the delay is just an approximated value

13. EXACT RDC v/s 1 bit RDC
IP header can not accommodate multi bit delay value
1 bit RDC , employs a CE bit in the IP header to accommodate the delay value

14. Using this simulation model 4 systems are compared:
1)Drop Tail
2)ECN
3)Exact RDC
4)1 bit RDC

15. Simulation

16. Layered Video Streaming
Hierarchically Encoded Video
Each layer requires same delivery bandwidth
Data of a layer is played only if subsequent lower layers are received
Additional layers enhances playback quality at the cost of more bandwidth

17. TCP as the Transport Protocol
Traditional TCP not suitable for real time applications i.e. streaming purposes
TCP connections introduce bandwidth and delay variations
Retransmission in TCP ensures reliable data communication but might induce playback latency in the system

18. Video Source Streaming Algo
Dynamic decision to add/drop a video layer
Decision is based on observed occupancy of the TCP sending buffer and the buffer size

19. Video Source streaming Algo

20. Simulation Different Simulation Cases -10 Stream Data Simulation
-Streams with Different Round Trip Time

21. Simulation-10 Stream

22. Simulation-10 Stream

23. Simulation-100 Stream

24. Simulation-Different Round Trip Time

25. Performance Summary
Exact RDC and 1Bit RDC performs better than the DropTail and ECN method
Exact RDC performs better than 1Bit RDC
1Bit RDC is easy to implement

26. Conclusion
TCP connections with Receiver-based Delay Control (RDC) works well for carrying layered video streams.
1-bit RDC do not use accurate delays
Additionally, a method for streaming hierarchically-encoded layered videos over TCP connections.

27. Conclusion RDC connection - utilize available bandwidth
-reduce playback latency
-friendly to competing TCP connections
Combination of our RDC and layered video source algorithms offers an attractive approach to streaming video over IP networks.

28. Our Suggestions on the Paper
The paper is written in a lucid language and is very easy to understand. There are lots of simulation results which makes understanding better.
The paper describes the streaming of video using an algorithm that is only simulated not implemented in real world. So we can’t be very sure of the success rate of the algorithm.

29. Thank You for your patience