Optimization and early-warning in DSL access networks based on simulation

Abstract

Network providers operate large DSL-based access networks to offer customers Broadband Internet. These networks are observed and managed by Performance Management Systems (PMS), that capture the actual situation to support network administration. In this regard, the administrator can cope with incidents such as link failures or congestion. We present an application for optimization and forecast of traffic distributions in DSL networks as an addition to an existing PMS. This application makes heavy use of simulation. In this way, we give a description of traffic models based on real network performance data reflecting: (I) individual subscribers and (II) an aggregated model for multiple subscribers. Then, we introduce the overall simulation approach based on the Network Security Simulator NeSSi2. The evaluation takes place by a use case for simulation-based verification of applied optimization strategies and a use case for continuous forecast to predict upcoming link congestion.