Deterministic Network Slice Instance Policy for Intra and Inter Slice Resource Management in 5G

Network slicing with Software Defined Networks and Network Function Virtualization enhances the network operator's serviceability, resource availability and coverage capability in 5G networks. The slice instance policies play a prominent role in addressing the infrastructure business rules, ownership policies and priorities of network operators, and service providers in multi-tenant scenarios. In this regard, the proposed Deterministic Network Slice Instance Policy model considers the slice instance policies, deterministic rules and dynamic programming approach to allocate and schedule the resources, and form the deterministic network slice instances. The slice instance policies incorporate slice isolation and customization using dedicated and shared resource allocation. The deterministic dynamic programming approach maximizes the resource allocation and throughput rate, and minimizes the cost and queue waiting time in resource scheduling. The proposed Deterministic Network Slice Instance Policy using Deep Reinforcement Learning considers the policies (slice instance), rules (deterministic), environment (states and rewards) and feedback in the deep reinforcement learning approach to enhance the performance of the network slice. Simulations indicate that the proposed model improves the resource allocation and data flow rates, and considerably reduces the average queue waiting time for slice instances.