Title:Federated Orchestration for Network Slicing of Bandwidth and Computational Resource

Abstract:Network slicing has been considered as one of the key enablers for 5G to support diversified IoT services and application scenarios. This paper studies the distributed network slicing for a massive scale IoT network supported by 5G with fog computing. Multiple services with various requirements need to be supported by both spectrum resource offered by 5G network and computational resourc of the fog computing network. We propose a novel distributed framework based on a new control plane entity, federated-orchestrator , which can coordinate the spectrum and computational resources without requiring any exchange of the local data and resource information from BSs. We propose a distributed resource allocation algorithm based on Alternating Direction Method of Multipliers with Partial Variable Splitting . We prove DistADMM-PVS minimizes the average service response time of the entire network with guaranteed worst-case performance for all supported types of services when the coordination between the F-orchestrator and BSs is perfectly synchronized. Motivated by the observation that coordination synchronization may result in high coordination delay that can be intolerable when the network is large in scale, we propose a novel asynchronized ADMM algorithm. We prove that AsynADMM can converge to the global optimal solution with improved scalability and negligible coordination delay. We evaluate the performance of our proposed framework using two-month of traffic data collected in a in-campus smart transportation system supported by a 5G network. Extensive simulation has been conducted for both pedestrian and vehicular-related services during peak and non-peak hours. Our results show that the proposed framework offers significant reduction on service response time for both supported services, especially compared to network slicing with only a single resource.