Bandwidth allocation for self-similar traffic consisting of multiple traffic classes with distinct characteristics
Peggy Orenstein, HoSoo Kim, Christopher Lau · 2002
It is now generally accepted that modern network traffic (for example, Ethernet, VBR and Web) is 'self-similar' in nature, i.e., the statistical characteristics of the traffic stream remains largely unchanged when examined at several different scales. One appropriate model for self-similar traffic is fractional Brownian motion (FBM). The main contribution of this paper is to formulate and verify a set of general rules that can be applied to calculate the FBM parameters of an aggregate traffic profile consisting of multiple input streams, each with its own characteristic burstiness profile. The parameters of the aggregated FBM traffic can then be used to calculate the effective bandwidth of an aggregate traffic pattern formed by multiplexing a mixture of heterogeneous traffic streams.