Conformance analysis in networks with service level agreements
Computer Networks and ISDN Systems
| Hi-index | 0.00 |
We introduce the concept of generalized Stochasti-cally Bounded Burstiness (gSBB) for Internet traffic,the tail distribution of whose burstiness can be boundedby a decreasing function in a function class with littlerestrictions.This new concept extends the concept oStochastically Bounded Burstiness (SBB) introduced byprevious researchers to a much larger extent, -- whilethe SBB model can apply to Gaussian self-similar inputprocesses, such as fractional Brownian motion, gSBBtraffic contains non-Gaussian self-similar inputprocesses, such as 驴 -stable self-similar processes which arenot SBB in general.We develop a network calculusfor gSBB traffic.We characterize gSBB traffic by thedistribution of its queue size.We explore property ofsums of gSBB traffic and relation of input and outputprocesses.We apply this calculus to a work-conservingsystem shared by a number of gSBB sources, to analyzethe behavior of output traffic for each source and totimate the probabilistic bounds for delays.We expectthat this new calculus will be of particular interest inthe implementation of services with statistical qualita-tive guarantees.