Abstract
Traffic modelling has always been the prior to numerous network engineering tasks like network planning or bandwidth yield management. It is still a huge research domain mainly based on advanced signal processing techniques. It can also help to better understand the underlying aspects of the usage of a telecommunication technology as well as to forecast its evolutions.
Key Terms in this Chapter
Ill-Posed Linear Inverse Problem: A linear inverse problem can be described by the equation: m = A d, where m is the observed data, A is a linear operator and d is the unknown parameter. When the dimension of the kernel of A is greater than zero, the problem is said to be ill-posed.
Underlying Traffic Structure: New components of traffic stemming from principal component analysis tool constitute the underlying traffic structure of a network. This corresponds to a new basis of traffic trends with a lower dimension that sum up all shapes of traffic evolution observed in a network.
Traffic Profile: A traffic profile is a sequence of measures over a specific period of time. It can be the traffic profile of a flow or a link count.
Cyclostationnary: A traffic measure is called to be cyclostationnary if it repeats the same cyclic shape within time. The most important part of the IP traffic is cyclostationnary as it reflects the human periods of activity (daily and weekly period).
Statistical Multiplexing: Statistical multiplexing is a method of making the most efficient use of the bandwidth available. Different flows share the same static resource (e.g. the bandwidth of a central link) and the idea is to allocate the bandwidth to each flow in order to prevent peaks from occurring at the same time on all the flows, which would result in a packet loss.
Bursty: We say that Internet traffic is bursty in order to explain that the bandwidth used by an IP flow is very irregular. It can go from 1 Mb to 10 Mb in one second. It is closely linked to the concept of sporadicity, which is the ratio of the peak rate and the mean rate of a flow. It reflects the on-off nature of individual connections.
Long Range Dependency: A self-similar phenomenon behaves the same when viewed at different scales on a dimension (space or time). Self-similar processes are said to exhibit long-range dependency.