Characterization of Background Traffic in Hybrid Network Simulation
Abstract
The information transfer across the battlespace is expanding at an ever increasing pace. Simulations have to be carried out to ensure the quality of service of the different applications (sensors, command and control systems, voice communications, ...) all integrated in a dynamic network environment. Two approaches are common: discrete event simulation and fluid approximation. A discrete event simulation generates a huge amount of events for a full-blown battlefield communication network resulting in a very long runtime. The usability of this simulation technique is very limited for a rapid changing combat theater. A faster fluid based approximation lacks the accuracy wanted for a realistic simulation. A hybrid simulator separates the traffic in two classes. The packets of the foreground traffic, for which fine-grained performance details are needed, are simulated by an event driven approach, while the background traffic, for which less detailed information is required, is approximated by a fluid model. A research area of this hybrid approach is the characterization of the fluid data flow. Classical statistics are inappropriate for real network data due to the heavy-tailed behaviour and the self-similarity of network traffic. This paper discusses the estimates based on the Large Deviations asymptotic, the Central Limit Theorem and a range of scaling laws in between both limits, a so called Moderate Deviations approximation. Based on real network traces the loss probability of a data stream can be estimated for all models. The results of this protocol independent analysis can be incorporated in a discrete-event simulation allowing a considerable reduction of computation time and memory consumption. The desired fine-grained details of the foreground application behaviour can be obtained while taking into account the background traffic.
Document Details
- Document Type
- Technical Report
- Publication Date
- Oct 01, 2006
- Accession Number
- ADA478802
Entities
People
- Antoine Van De Capelle
- Bart Scheers
- Ben Lauwens
Organizations
- Royal Military Academy