2nd International ICST Conference on Performance Evaluation Methodologies and Tools

Research Article

A Histogram-Based Stochastic Process for Finite Buffer Occupancy Analysis

Download543 downloads
  • @INPROCEEDINGS{10.4108/valuetools.2007.1972,
        author={Enrique Hernandez-Orallo and Joan Vila-Carbo},
        title={A Histogram-Based Stochastic Process for Finite Buffer Occupancy Analysis},
        proceedings={2nd International ICST Conference on Performance Evaluation Methodologies and Tools},
        proceedings_a={VALUETOOLS},
        year={2010},
        month={5},
        keywords={Traffic modeling Network QoS Stochastic Analysis},
        doi={10.4108/valuetools.2007.1972}
    }
    
  • Enrique Hernandez-Orallo
    Joan Vila-Carbo
    Year: 2010
    A Histogram-Based Stochastic Process for Finite Buffer Occupancy Analysis
    VALUETOOLS
    ICST
    DOI: 10.4108/valuetools.2007.1972
Enrique Hernandez-Orallo1,*, Joan Vila-Carbo1,*
  • 1: Departamento de Informatica de Sistemas y Computadores Universidad Politecnica de Valencia Valencia, Spain
*Contact email: ehernandez@disca.upv.es, jvila@disca.upv.es

Abstract

This paper proposes to use histograms for characterising network traffic and a simple stochastic process for network performance analysis. The result of this process is the buffer occupancy histogram (queue length distribution) using a finite queue model. From this buffer occupancy histogram we detail how to obtain another interesting performance parameters like cell loss ratio and network delay distribution. The proposed method has been extensively evaluated using real traffic traces. These evaluations show that the model is accurate. Applications of this model are very wide: analysis and prediction of QoS parameters, network dimensioning and provisioning, traffic admission control, etc.