Decentralized Scalable Dynamic Load Balancing among Virtual Network Slice Instantiations

In the virtualized environment of 5G networks, the control and management of dynamic network slices poses a set of challenges that are still largely unsolved. Though the architectural framework and the elements of abstraction and orchestration mechanisms have been defined, the dynamic orchestration of resources based on them entails the adoption of existing sophisticated control techniques, or the design of new ones for the specific environment. In the present paper, we address the problem of load balancing among multiple network service chains (which represent network slice instantiations of a Network Service Provider referring to a specific vertical application) originating from different Points of Presence (PoPs). For scalability reasons, we want to maintain the problem within an informationally decentralized setting, where each PoP has the knowledge of the aggregate workload generated by the slice users accessing through it, but not of that of the other PoPs (to avoid the exchange of information for control purposes). By taking also into account power consumption policies of the Infrastructure Provider, we find a set of candidate team-optimal solutions to this load-balancing problem, which are characterized by piecewise-linear functions, and compare their performance with that of other resource allocation strategies.