Traffic Analysis

Traffic analysis is concerned with analysis of the load to be handled by the system. This may be difficult to analyse as many different forms of media may be handled simultaneously and be distributed across the system.

Traffic characterisation and performance evaluation of networks (that support the system) are essential to providing a guaranteed service: optimising and dimensioning the network resources (i.e Buffer - space below - and bandwidth) is also important. Fortunately, the multimedia handled is quantifiable and describable and hence predictable.

A variety of methods have been developed to model Multimedia traffic.

The reader is referred the Chapter 7 of Networked Multimedia Systems by Raghavan and Tripathi for full details of this type of analysis. Here we simply list the methods available with a brief description:

Markov Modulated Poisson Process (MMPP)

-- A window based time series method for estimated traffic.

On/Off Models

-- Designed to capture the active/idle behaviour of a typical data source. Sometimes referred to as a packet-train model.

Fluid Models

-- Models traffic as a fluid flowing through a system -- appropriate in ATM-based networks, not for compressed video.

Time-Series Models

-- These type of models are especially appropriate when dealing with compressed data streams.

Transform-Expand-Sample (TES) Model

-- Used to generate synthetic traffic streams and suitable for trace-driven simulations.