Fair Allocation of Cloud & Physical Resources for Multi-class Users

In Future Networks, network cloud computing is expected to bring a disruptive change in satisfying the users QoS/QoE requirements, through resource virtualisation [1]-[3]. However, two important technical challenges remain to be resolved first: a) isolation, to ensure high performance for certain users groups and applications, and (b) dynamic reallocation of cloud and physical resources according to traffic and user requirements. To address both of these challenges, it is necessary to jointly manage both the virtual and physical resources, in an efficient automated way, as well as to fulfill the individual user requirements. For example, certain users may simultaneously consume virtual resources via WiFi and physical resources via 3G, while other users are only assigned limited virtual resources that cannot fulfill their QoS. Existing work in the literature focuses on techniques for fairness optimisation in specific systems [1], [2], e.g. large-scale clouds. The major contribution of this work is the introduction of the H-Fairness function (Fairness for Heterogeneous Environments), for measuring fairness in both virtual and physical resources allocation among different user classes based on their individual requirements. The H-Fairness function extends the classical definition of fairness introduced by Jain [4]. The latter does not capture the requirements of the heterogeneous environment of future networks, as it cannot distinguish different user classes (e.g. gold, silver and bronze users) and resource types.