Tags:Decentralized learning, Fisher Market, Market Equilibrium, Network slicing, Resource allocation and Trading post mechanism
Abstract:
Network slicing (NS) is a potential technology to support a higher degree of heterogeneity and flexibility required by next-generation services in 5G networks. One of the essential elements in network slicing is resource allocation, which needs to ensure slice tenants the protection of their service level agreements (SLAs) while optimizing network efficiency. We propose a resource-sharing scheme based on the Fisher market model and the Trading post mechanism that can achieve efficient resource utilization while catering to multi-level fairness, dynamic load conditions and SLA protection. In the proposed scheme, each service provider (SP) is given a budget representing its infrastructure share or purchasing power in the market. The SPs acquire different resources by spending their budgets to offer the service to different classes of users, which are classified according to their service needs and priorities. We assume that SPs employ the $\alpha$ fairness criteria while delivering the service to their subscribers. The proposed allocation scheme aims to find a market equilibrium(ME) that provides allocation and resource pricing whereby each SP's needs are met, and resources are fully utilized. We show that the ME solution problem can be formulated as a convex optimization problem whose primal and dual optimal solution provides equilibrium allocation and pricing. We build a privacy-preserving learning algorithm enabling SPs to reach ME in a decentralized fashion. We theoretically evaluate the proposed allocation scheme's performance by comparing it with the Social Optimal and Static Proportional sharing schemes. Finally, we run extensive numerical simulations to assess the fairness and efficiency properties of the proposed scheme.
Fisher Market Model Based Resource Allocation for 5G Network Slicing