Capacity planning in IP Virtual Private Networks under mixed traffic

Abstract

A new mechanism for capacity planning of a Virtual Private Network, carrying both Quality of Service (QoS) and Best-Effort (BE) IP traffic is introduced. The link bandwidth allocation in the network is based on a hybrid combination of analytical and simulation models. The allocation tool works in three phases: phase 1 and 2 deal with the assessment of the capacities needed to satisfy requirements for QoS and BE traffic individually; the effects of multiplexing gains arising from traffic mixing are taken into account in phase 3. There are two analytical parts, based on a simplified version of call-level traffic models, derived from teletraffic theory of circuit-switched networks, and on an algorithmic procedure to compute the distribution of average values of BE traffic flows, respectively. The same algorithm is used inside a fast simulation model, adopting a fluid representation of TCP traffic aggregates. The tool is tested both on simple networks and on a more complex real national backbone.