A statistical analysis of network parameters for the self-management of lambda-connections

Abstract

Network monitoring plays an important role in network management. Through the analysis of network parameters (e.g., flow throughput), managers can observe network behavior and make decisions based on them. The choice of network parameters although should be relevant for each specific objective. In this paper, we focus on the analysis of network parameters that are relevant for our self-management of lambda-connections proposal. This proposal consists of an automatic decision process to offload large IP flows onto lambda-connections. This paper aims at statistically analyzing a list of potential network parameters as relevant estimators for flow volume. The main contribution of this work is the introduction of a statistical methodology to validate that some few network parameters can be considered as good predictors for flow volume. These predictors are therefore of great interest to be used in our automatic decision process. © 2009 IFIP International Federation for Information Processing.