Framework for Wireless Networks Management based on Network and Traffic Modeling with QoS Objectives

The highly erratic and unpredictable behavior of wireless network traffic can be attributed to the combination of different factors: the variable number of users, the capacity of the access links/points, traffic matrices, users’ mobility, background traffic and the diversity of applications, services and user behaviors. These issues, together with the increasing demand for bandwidth and the every day appearance of new applications and QoS requirements, have been driving a strong need for tools that can aid in the global management of mobile networks. This extended abstract proposes a management framework for wireless networks that integrates a wireless network model and a set of procedures allowing (i) to predict network Quality of Service (QoS) evolution for varying number of users and user characteristics and (ii) to assess the need for network upgrades. The availability of this framework can help network managers and service providers to plan short-term or long-term network reconfigurations and upgrades or design new strategies for network management, traffic routing, service provisioning and other critical network operational issues. The correct planning and location of network resources can greatly increase network operation efficiency and optimize network QoS parameter values. The framework’s inherent wireless network model [1], [2] incorporates all network important functional characteristics and is able to predict QoS parameters based only on traffic that enters and leaves the network at its different access points (APs) (the inputs of the model), without requiring any a priori information on the network structure, traffic routing or users mobility profiles. The network model is built from a set of (past) measurements of the APs inbound/outbound traffic and the corresponding network QoS metrics. After being inferred, the model can predict the network QoS parameters based on traffic predictions obtained from previously inferred traffic source models and based on an estimation on the future number of users (and their respective profiles). Besides, the network model parameters can be mapped to real network parameters: traffic flows between network users and APs, users’ mobility characteristics, background traffic, network operational characteristics and inherent fixed QoS. The extended abstract is organized as follows: section II presents the framework wireless network model, including the main steps of its inference procedure; section III presents the proposed network management framework and section IV presents the network and simulation scenarios that will be used to demonstrate the usefulness and accuracy of the proposed framework.