The availability of good stochastic models of network traffic is the key to developing protocols and services. A precise statistical characterization of packet interarrival time, size distribution, and connection arrival rate help network engineers to design network equipment and evaluate its performance.
Key Terms in this Chapter
Multifractality: A generalization of self-similarity in which the small time-scale behaviour of the process shows local variations in the scaling parameter.
Heavy-Tailed Distribution: A statistical distribution with tails that decay subexponentially.
Hurst Parameter: A qualitative measure of selfsimilarity related to the scaling parameter a defined for long-range dependent processes.
Aggregation (of a time series): The action of averaging the time series over non-overlapping blocks of constant size.
Long-Range Dependence (also known as long memory): A property found in stochastic processes with strong low-frequency components.
Fractals: Objects (in particular, figures) that have the same appearance when they are seen on fine and coarse scales.
Self-Similarity: When applied to stochastic processes, it indicates that the process follows the same distribution on all time scales.