It is well-accepted that Pareto (heavy-tailed) and Gamma (short-tailed) distributions can be used to model the on-off time of a bursty traffic. If \(X\) is Pareto, i.e. \(X\sim\dfrac{ac^a}{(x+c)^{a+1}}\), and \(Y\) is Gamma, i.e. \(Y\sim\dfrac{y^{\alpha-1}e^{-y/\lambda}}{\lambda^\alpha\Gamma(\alpha)}\), then \(R=X+Y\) models the time between successive on-off cycles.
[more]
Phit and Abe (2006) Packet Inter-arrival Time Estimation Using Neural Network Models (IC'06)
This paper proposed a neural network model to estimate packet inter-arrival time. The neural network model is to take past \(n\) inter-arrival time into account to estimate the next arrival. Different model is proposed (such as linear or sigmod functions) and the parameters of the function is determined by machine...
[more]
Durand and Flajolet (2003) Loglog Counting of Large Cardinalities (ESA'03)
Assume we have a multiset (a set which each item can have multiplicity \(> 1\)) of
elements from a universe of infinite size. The number of elements in the
multiset is finite and we define the cardinality of the multiset to be the
number of distinct elements in it.
[more]
Lu et al (2008) Counter Braids: A Novel Counter Architecture for Per-Flow Measurement (SIGMETRICS)
Objective: Report per-flow throughput measurement using minimal memory so that it can be implemented using SRAM and applied to high-speed network.
[more]
Wu, Xiao, Li, Sun (2008) A Distributed Algorithm for Finding Global Icebergs with Linked Counting Bloom Filters (ICC'08)
Objective: A collection of objects are stored in multiple hosts distributively
such that each host only has a partial collection. The problem is to tell the
icebergs amongst these objects without using centralised server.
[more]