QoS技术基础-chpt3.ppt

Traffic Characteristics The Problem; Packet Voice Modeling; CBR model; Two-state model; Fluid-flow model; MMPP model; Video Traffic Characteristics and Modeling; Video traffic characterization; Packet video modeling; AR model; Fluid flow model; MMPP model; More Practical Model; Leaky bucket model; Token bucket model; Model used in ISPN; The Problem Traffic models play a significant role in the design and engineering of IP networks: Configuration of networks; Network QoS; Performance evaluation; Related issues; It is required to develop rather sophisticated traffic models to carry out accurate design, engineering and performance evaluation. CBR Model for Voice Without silence suppression, digitized voice (compressed or uncompressed) can be simply modeled as a bits stream with constant rate (i.e. the number of bits per second = constant); In this context, packetized voice stream can be roughly viewed as having constant packets rate: Fixed packet size; The number of packets per second = constant; CBR Model for Voice (cont’d) The multiplexing problem A simple multiplexing model CBR Model for Voice (cont’d) As for the multiplexing model, the followings need to be worked out: The buffer size, B; The average waiting time, E(W); The utilization ratio: To keep the system stable, the utilization level needs to be less than 1; Worst case analysis of the required buffer size: The buffer size of (N-1) packets, not including the one in service, will suffice to handle the case; Zero blocking probability; The survivor function of delay* (the length of time spent in the queue waiting for service) CBR Model for Voice (cont’d) As for nD/D/1 system, the F(x): Not in a simple and straightforward form; For details, refer to the following article: Eckberg, Jr., A. E., “The Single Server Queue with Periodic Arrival Process and Deterministic Service Times,” IEEE Trans. On Communications, COM-27, 3(March 1979):556-562. Example and comparison: N = 5; r = 59808 cells/sec; C = 312500 cells/s