1. A Framework for the Analysis of Adaptive Voice over IP Casetti, C.; De Martin, J.C.; Meo, M. Communications, 2000. ICC 2000. 2000 IEEE International Conference on

2. Outline n Introduction n Network-driven end-to-end control over variable bit-rate voice sources n Analytical approach n Case study n Conclusions

3. Introduction n Voice communication has received special attention and the possibility to offer telephony services. n How to best deliver voice over IP networks? n Adaptive Voice over IP –Solutions based on variable bit-rate speech coders that adapt to network conditions.

4. Network-driven end-to-end control n The main goal of the algorithm –to reduce the load on the network when queue build-ups occur.

5. Network-driven end-to-end control n Variable-rate adaptive schemes for multimedia streams in the past –analyze the behavior of a buffer control mechanism. [Cox and Crochiere] –an end-to-end feedback control (the involvement of intermediate nodes is required) [Bially] –based on the state of network [Bolot and Turetti] –a sender-based packet loss rate estimation scheme [Sisalem and Schulzrinne] –an analytical model for an end-to-end feedback control mechanism [Yin and Hluchyj]

6. Network-driven end-to-end control n A feedback can be expressed as aggregate rate, traffic delay or loss rate at the receiver. n The effectiveness of the control is strongly related to the delay-load relation. n The rate control algorithm uses the information carried by cyclic RTCP receiver reports to let the source know the state of the on-going connection.

7. Analytical Approach n Model of the source in isolation n Model of the Network n Interaction between a source and the network

8. Analytical Approach n The basic assumptions –every source is modeled in isolation and is statistically independent from the other sources. –The interaction of the sources with the rest of the network is perceived through the feedback information conveyed back to each source. –The model dynamics are determined by a number of driving stochastic processes: new call generation call termination source bit-rate increase source bit-rate decrease

9. Model of the source in isolation n The user’s dynamics –idle –talking –silent n the end-to-end control mechanism –N different bit rates ( B 1 ~B N ) –states: I ( l, m ) where I t s λ/2 μμ β α

11. Model of the network n The model of a network with K sources n the state of the network n the state space for the network

12. Model of the network n The probability that the network is in state n n The aggregate generated traffic T(n) n the delay experienced by the packets when the system is in state n

13. Interaction between a source and the network n The probability that the network is operating in critical conditions –where n The probability that the network is operating in underload conditions –where

14. Interaction between a source and the network n The round trip time is twice the packet delay in state n, 2*D(n) n Evaluating the reaction rate as the inverse of the reaction time, we can derive

15. Case study n Three bit-rates –8 kb/s (e.g., ITU-T G.729) –13 kb/s (e.g., ETSI GSM-EFR) –64kb/s (e.g., ITU-T G.711) n P b : the probability of being in one of the states where the system is losing packets where

16. Case study

21. Conclusions n We propose a framework for the analysis of adaptive, variable bit-rate sources operating in a packet network environment. n It can be useful in the preliminary phases of study of complex systems, for comparisons between different control algorithms of for tuning the parameters of the algorithm.