Abstract

Scalability problems arise when the Real Time Control Protocol (RTCP), which is the control protocol of the Real-time Transport Protocol (RTP), is used in large multicast groups. The problems include: increased feedback delay, increased storage state at every member, and ineffective RTCP bandwidth usage.

We have designed a Scalable RTCP (S-RTCP) which is based on a hierarchical structure in which members are grouped into local regions. For every region, there is an Aggregator (AG) which receives the feedback Receiver Reports (RRs) sent by its local members. The AG summarizes important information in the RRs, derives some statistics, and sends them to a Manager. The Manager performs additional statistical analysis to monitor the transmission quality and to identify regions suffering massively from congestion.

A simulation of S-RTCP using the Network Simulator (NS) showed that S-RTCP alleviates some of the RTCP scalability problems and makes effective use of RRs. Consequently, the feedback provides timely and useful QoS information required for network monitors and for adaptive applications. Details of the simulations and performance analysis are presented and described which show the advantages of using S-RTCP over the original RTCP.

Keywords: Real-time transport protocol (RTP); Real time control protocol (RTCP); Scalability

Article Outline

1. Introduction

2. Overview of RTCP and its scalability problems

2.1. RTCP scalability problems in large multicast groups

2.1.1. Feedback delay

2.1.2. Increasing storage state

2.1.3. Multicasting RRs to the whole group

2.1.4. Initial feedback flood

2.1.5. Bye flood

3. The scalable RTCP scheme

3.1. Overall view

3.2. Scheme entities

3.3. Scheme description

3.3.1. Selection of Parent AG and formation of local regions

3.3.2. AG leaving or crashed

3.3.3. SR and Bye RTCP reports

3.4. Contents of an AGR

3.5. Functionality of the Manager

4. Benefits of the scheme

5. Performance of S-RTCP: simulations and evaluations

5.1. General setup of the simulations

5.2. Simulation results

5.3. Storage

5.4. Feedback delay and frequency of RRs

5.5. Bandwidth utilization

6. Related work

7. Discussion and future work

8. Conclusion

Acknowledgements

References