Research On Key Technology Of Cloud Service Scheduling For IoT Devices

The continuous development of Internet of things technology is embodied in the popularity of IoT devices and connected with the Internet.IoT devices are limited by battery capacity,storage space,computing power and other factors that make them unsuitable for performing complex tasks alone.Cloud computing technology provides the platform and service support for the expand of IoT devices.On the one hand,IoT devices need to use cloud resources to expand their capabilities,on the other hand,through cloud platform,users can also use the services provided by IoT devices.This paper summarizes three cloud service modes for IoT devices and explores how cloud platforms can use virtualization technology more effectively to provide services for IoT devices.The goal of this paper is to build a cloud service platform for IoT devices based on virtualization technology.The main work and innovative achievements in this paper are as follows:1)For the service request mode,this paper studies the problem of the optimal virtual machine placement for IoT devices.A virtual machine placement algorithm to minimize round-trip time(RTT)is proposed to reduce the delay of device service requests.In this service mode,the service program running in the virtual machine provides services for IoT devices.First of all,this paper proposes a virtual machine dynamic placement framework.In this framework,the network topology of the physical machine in cloud center is constructed by Fat-tree.The service requests of IoT devices are distributed to the corresponding core switches through load balancing modules.In this paper,the round-trip time between core switches and physical machines is used as a metric of service delay,then a virtual machine placement algorithm for minimizing round-trip time is proposed.At the same time,taking into account the failure of core switches,a virtual machine rescheduling algorithm is proposed to reduce the average RTT fluctuation of all service requests for ensuring the stable operation of IoT devices.Finally,we use CloudSim to verify the algorithm proposed in this paper,and experimental results show that the algorithm proposed in this paper can effectively reduce the service request delay for IoT devices.2)For the task offloading mode,this paper studies the problem of the task offloading for IoT devices.A task offloading strategy based on two-layer reinforcement learning algorithm is proposed to trade off between the cloud resource utilization and delay.Theoretical analysis shows that in the process of task offloading,it is often contradictory to improve the utilization of cloud resources and reduce the task delay simultaneously.In this paper,we use reinforcement learning algorithm to model the task offloading problem.Then,the weighted return value is introduced,and the weighted parameter is adjusted to trade off between the cloud resource utilization rate and delay.In the reinforcement learning algorithm,the high dimension state space and action space make the algorithm converge slowly.Therefore,a two-layered reinforcement learning algorithm is proposed to solve this problem.First,the k-NN algorithm is used to divide the physical machines into K clusters.The physical machines in each cluster have similar bandwidth and waiting time.The first layer of the algorithm is implemented by deep reinforcement learning,which determines the optimal cluster of physical machines according to the current offloaded task.The second layer is implemented by Q-learning algorithm,in which the optimal strategy is learned in each cluster to select the physical machine for running the offloaded task.Finally,simulation experiments show that algorithm can effectively trade off between cloud resource utilization and task delay by adjusting the weight parameters of the return value.At the same time,the proposed two-layered reinforcement learning algorithm can learn the optimal task offloading strategy faster.3)For the smart device as a service mode,this paper studies the problem of dynamic service composition for IoT devices and proposes an adaptive web services composition algorithm based on subtask-decomposition strategy.The algorithm will learn jointly by considering the association between different service composition tasks.First,the Markov decision process(MDP)is used to model the service composition problem where each action corresponds to an alternative service.Its goal is to learn the best service composition strategy to maximize the QoS of combined service.Secondly,in order to consider the association between different service composition tasks,complex service composition tasks are decomposed into multiple independent subtasks based on graph theory.Then,the relationship between common subtasks belonging to different service composition tasks is established,and a subtask-assistance updating strategy is proposed.Based on this strategy,a multi-Q learning algorithm is proposed,which is suitable for large-scale dynamic service composition problems.Finally,the algorithm is verified by QWS dataset,the results show that the proposed algorithm has faster convergence speed under dynamic environment.At the same time,the speed of convergence of tasks with the same subtasks can be speeded up by using subtask-assistance updating strategy.4)This paper proposes a connecting framework of IoT devices based on linux container technology and builds cloud platform for IoT devices combined with the first three research contents.Firstly,this paper describes the operation mechanism of the proposed connecting framework of smart device in detail and Compared with the existing frameworks,the proposed framework is more convenient to be connected and more scalable.Then it describes in detail the cloud platform for IoT devices based on the connecting framework proposed in this paper,which implements the three service modes provided by cloud center for IoT devices.Finally,the intelligent community oriented health monitoring system and the health care platform of USTB are developed by using the cloud platform for IoT devices.And as a demonstration project of intelligence old-age care for the Ministry of civil affairs,the system and platform are applied in University of Science and Technology Beijing community and Two Li Zhuang community.