Fairness-Aware Online Slicing and Adaptive Pricing in Multi-Tenant Computing Networks

The multi-tenant computing network (MCN) is an information infrastructure that integrates ubiquitous computing resources to enable flexible, efficient, and reliable computing resource allocation. Network slicing is a key virtualization technology in MCN that enables efficient sharing and allocation of virtual networks. In online settings, tenant arrivals are unpredictable and allocation decisions are irrevocable. Existing methods often struggle to balance fairness and efficiency due to their offline assumptions, complexity, and reliance on privacy. To address these issues, we study the problem of online slicing fair allocation and dynamic pricing in MCN. First, we consider an online Fisher market model where each tenant's budget and the utility derived from the tenant's service subscribers are revealed online. Next, we design aFair, efficient, andPrivacy-preservingonlineSlicing algorithm FPoS that dynamically adjusts resource prices using only observations of tenants' consumption behavior (i.e., their revealed preferences). This approach allows the platform to approximate market equilibrium without accessing tenants' private parameters, such as their budgets or utility functions. FPoS is based on the posted-price mechanism and uses a multiplicative rule to update prices, achieving real-time fair allocation while avoiding dependence on private parameters of tenants. We derive the expected regret upper bound as$\mathcal {O}(\sqrt{N})$of FPoS through regret analysis. Experimental results show that FPoS outperforms other online baseline methods in multiple metrics of fairness and efficiency, and its overall performance is closest to the offline optimal solution.