Resource Scheduling Algorithms For Delay Optimization In Mobile Edge Computing

With the development of the Internet of Everything,many new applications such as driverless and intelligent video has emerged rapidly and meanwhile there arouse stricter demands for delay.Therefore,mobile edge computing was proposed,which can effectively solve the problems of delay and limited battery capacity.Research on mobile edge computing lies in the areas of the system architecture,resource management,green energy conservation and security privacy.Among them,the area of resource management is the most relevant to the computer science research,and has also been paid much attention by many researchers.In view of the high latency problem of traditional cloud computing using WAN transmission,the mobile edge computing provides users with low-latency,close-range cloud services via deploying servers at network edge nodes.Due to the distributed deployment of servers and the limited resources of each node for the mobile edge computing,the existing resource scheduling strategies can't work well.Moreover,it will bring much burden for the traditional resource scheduling on the adaptation into the novel network structure because the heterogeneous and coupled computing,communication and cache resources.Therefore,it is necessary to propose flexible resource allocation strategies to reduce the delay of data transmission.In view of the challenges listed above,we studied the resource scheduling algorithms for delay optimization in mobile edge computing.The main content and contributions are listed as follows:Firstly,concerning the computing resource scheduling problem in multi-server single-user systems,the Markov approximation algorithm is designed to achieve more efficient system computing resource scheduling.We construct the mathematical model.Then,we exploit the task allocation decision and expanding the computing capacity in the devices.Besides,the minimization of trade-off between delay and energy consumption is considered as the optimization target.And the Markov chain execution state is implemented on all feasible configurations by using the Markov approximation algorithm to achieve an approximate optimal solution in a relatively short period of time.We carry out simulations through C++,the results of which demonstrate that the proposed algorithm could effectively generate approximate optimal solutions and is superior to other traditional benchmark algorithms under various parameter settings.Secondly,to deal with the difficulty in the process of joint optimization of the communication and computing resources in multi-server and multi-user systems,we propose a distributed transmit power optimization algorithm based on the Markov approximation.Based on Shannon's theorem and link transmission characteristics,strategies for minimizing power consumption of users are modeled.By leveraging the Log-Sum-Exp approximation function,the original problem turns into a relevant approximated one.The self-adjustment mechanism of the distributed device is proposed to solve the approximation problem to obtain efficient solutions.Experimental results show that the proposed algorithm ensures fast convergence of transmit power and reduces user delay and energy consumption effectively.Thirdly,we design the collaborative cache mechanism to reduce the backhaul network congestion for the joint optimization of the communication,computing and cache resources in multi-server systems.Meanwhile,we utilize the deep reinforcement learning to optimize user delay.Besides,based on the dynamic change of mobile edge computing environment,a server-based joint optimization framework of collaborative caching and computing is introduced to realize dynamic allocation of multiple resources in the community.The deep Q network algorithm in the deep reinforcement learning is exploited to optimize cumulative delay and meet user delay requirements.Simulations via the TensorFlow framework and Python programs illustrate the effectiveness,reliability and adaptability of the proposed scheme.