A Game Theoretic Approach for Distributed Resource Allocation and Orchestration of Softwarized Networks

Softwarization of networks allows simplifying deployment, configuration and management of networkfunctions. The driving force towards this evolution is represented by Software Defined Networking (SDN)that allows more flexible and dynamic network resource allocation and management. Efficient resourceallocation and orchestration are two primary targets of this softwarization process; however, centralizedmethodologies result complex, and exhibit scalability issues. So, distributed solutions are to be preferredbut, in order to be effective, should quickly converge towards equilibrium solutions. In this paper, we focuson making distributed resource allocation and orchestration a viable approach, and prove convergenceof the relevant mechanisms. Specifically, we exploit game theory to model interactions between usersrequesting network functions and servers providing these functions. Accordingly, a two-stage Stackelberggame is presented where servers act as leaders of the game and users as followers. Servers have conflictinginterests and try to maximize their utility; users, on the other hand, use a replicator behavior and try toimitate other users decisions to improve their benefit. The framework proves the existence and uniquenessof an equilibrium, and a learning mechanism to converge to such equilibrium is proposed. Numericalresults show the effectiveness of the approach.