Quality of service driven resource allocation in network slicing

Abstract

Network slices are counted as key technology enablers to provide tailored services and isolation for different Fifth Generation (5G) cellular network application instances. A network slice is an independent end-to-end logical network over a shared physical substrate network, capable of providing a specific set of negotiated services to the customers, quite analogous to virtual machines offered in the domain of cloud computing. In this paper, we perform a systematic study of allocation and dynamic on-demand provisioning of network resources for the slices. The core parameters of Quality of Experience (QoE) to end-user systems, Network performance, and Operating efficiency are carefully applied while placing network virtual functions and determining the nodes, links, and resources for assignment to these slices. Network slices are expected to provide end-to-end capabilities while providing customized offerings. Slices can be further divided into various categories such as RAN, Transport and Core network slices. Our work examines the allocation of the above categories as independent sub-problems and latter amalgamate it into an endto-end multi-objective constrained optimization problem. Our proposed approach is influenced by Multiple Attribute Decision Making, Analytical Hierarchy Processing for slice assignment and enhanced Dinic’s Maximum Flow Method to find all possible virtual paths for allocations. Simulations are carried out through NS3 and results are compared against well-known algorithms