A client-based QoS approach using generalized extreme value theorem in multi-hop network environments

Client-side QoS monitoring allows mobile clients to participate in the service evaluation process. Through this approach, mobile clients play proactive role to collect and measure network parameters (e.g., delay, bandwidth) and report any potential service degradation back to the service providers. Adopting this model provides a major enhancement to the service evaluation process by allowing service providers to tune their network resources for a better service performance. In this paper, we present a novel approach that evaluates service performance through collecting service measurements from mobile clients (MCs) in a multi-hop network environment. The proposed approach utilizes a System of Linear Equations (SLE) and Generalized Extreme Value Theorem (GEV) techniques to predict network link performance degradation by estimating the delay extremes on each hop of a given network topology. Consequently, service performance can be evaluated and improved through a continuous assessment process of the network´s links behavior over time. Our experimental analysis indicates that our proposed approach provides better prediction of the link delays.