Energy Minimization for Distributed Microservice-Aware Wireless Cellular Networks

With the rapid development and widespread deployment of Internet of Things devices, existing networks face significant challenges in meeting the demands of emerging large-scale applications. In this paper, we propose a novel paradigm to address these challenges by decomposing large applications/services into lightweight microservices (MSs) distributed among small base stations (SBSs), each responsible for specific functions. Upon receiving a service request, a macro base station (MBS) invokes a series of SBSs that cache the required MSs to execute the associated computational tasks. The computed results are then returned to the MBS, which integrates and delivers the final result to the user. Under this framework, we investigate the joint problem of MS caching, computation task assignment, and computing resource allocation, aiming to minimize the total energy consumption. Various practical constraints such as users' latency requirements and the limited caching and computing resources of SBSs are taken into account. To facilitate the analysis, we transform the original minimization problem into an equivalent problem focusing on MS computation task assignment and computing resource allocation, which remains NP-hard. To tackle this challenge efficiently, we devise a two-stage method. In the first stage, we derive a closed-form expression for the computing resource allocation policy based on the MS computation task assignment. Subsequently, we introduce a two-side swapping oriented approach to explore an improved MS computation task assignment strategy. In addition, we propose the use of exhaustive and simulated annealing algorithms to approach the optimal and near-optimal solutions, respectively. Extensive simulation results demonstrate that our proposed algorithm achieves close-to-optimal performance and outperforms benchmark schemes significantly.