Research On Scheduling Algorithm For Green Mobile Edge Computing Systems

As a new computing paradigm,the remote cloud computing suffers from the long transmission delay and high cost of spectrum resource.To cover these shortage,mobile edge computing has attracted more attention from both academic and industrial researchers in recent years.In mobile edge computing systems,network edge devices,such as base stations(BS),are endowed with computing functionalities that are similar to the remote cloud.Therefore,they can serve as small data centers to deal with computing requests from users.Since the computing resources become closer to user terminals,mobile edge computing makes it possible to provide computing services with low latency and location-aware data processing capability.On the other hand,computing tasks usually require a large amount of energy consumption.While using renewable green energy to supply the system is both economic and carbon-free.However,the collection of green energy relies on the energy harvesting technique,which can be highly intermittent and unpredictable.If every BS works alone on mobile edge computing as well as energy harvesting,its performance would be limited both by computing capability and current energy supply of a single BS.This makes it difficult to make full use of mobile edge computing system.To provide better mobile edge computing services for users,corporations should be made among BSs by scheduling.In our energy harvested mobile edge computing network,apart from the spatial domain coupling caused by traditional geographical load balancing technique,energy harvesting also leads to a stringent energy constraint that couples the GLB decisions across time.This adds more challenges to the design of our scheduling algorithm compared to the existing works.Here,we apply the Lyapunov optimization technique with perturbation combining the geographical load balancing method to propose a new data scheduling algorithm that is suitable for our system.Our work mainly contains the following two parts:We design a new GLOBE algorithm,which can minimize the long time average cost of the system.With the use of Lyapunov optimization technique,this algorithm can work on single time slots without the request of future information of the system.Therefore,the GLOBE algorithm can work online.Also,we prove by theoretical analysis and system simulation that GLOBE algorithm makes a compromise between the system performance and the battery capacity.Furthermore,it performs much better than other baseline algorithms.We also provide a distributed implementation of the GLOBE algorithm.By using quad regularized Lagrangian dual decomposition algorithm,we can relax the constraints caused by the computing capability and decouple in the spatial domain.Each BS is enabled to make its own GLB decisions.The distributed algorithm lowers the complexity of the problem as well as improves its scalability and practicability when the system is at a large scale.Moreover,we verifies its effectiveness through simulation.Above all,our work proposes both centralized and distributed scheduling policies for the green mobile edge computing multi-BS network,and verifies their good system performance both theoretically and empirically.