Research Of Coupling Resource Allocation And Task Offloading Based On Fog Computing

In the recent years,with the rapid development of cloud computing,the collaboration between the Internet of Things(Io T)and cloud has become a trend and sensor-cloud has appeared.Open,flexible,and configurable service platforms for different types of applications are available by using the cloud platform for information processing and storage.Therefore,users can use or control virtual nodes provided by the cloud server to meet service requests,instead of purchasing or deploying a sensor network.However,conflicts will happen if a sensor receives commands from multiple users at the same time in the system.The main problems caused by the coupling problem are as follows: physical node request deadlock,long waiting time of user and low resource utilization.The characteristics of fog computing are local deployment,proximity to users,and low latency.By transferring user' requests from the cloud to the edge for management,the physical node resources can be efficiently and directly managed.In general,this thesis investigates the coupling resource allocation and task offloading strategy in sensor-cloud system.Then,the collaboration between the Io T and cloud is proposed to improve resource utilization and reduce the waiting time for users,which would make effective use of the fog computing resources.Therefore,this thesis provides new ideas to the coupling resource scheduling and offloading problems for future research.The main work of this thesis is listed as follows:(1)In order to solve the problem of coupling resource scheduling in sensor-cloud,a low coupling method based on fog storage is proposed.By designing the cache queue in the fog layer,priority is given to the nodes which are set to high priority firstly so that delay-sensitive tasks can be processed preferentially.Then,priority is given to the nodes with less computing resources,thereby effectively reducing the delay of subsequent services.In this thesis,a mathematical model is established for the coupling problem of fog storage and the problem is formally defined,which is proved to be NP-hard.Through simulation,several scheduling algorithms and the proposed low-coupling method of sensor-cloud based on fog storage are performed.In contrast,the experimental results show that the proposed algorithm can significantly reduce the number of scheduling rounds and slower response time.(2)In order to solve the problem of low utilization of physical nodes and low resource dispersion,a low coupling method of sensor-cloud based on double buffer queue is proposed.This thesis focuses on how to schedule the request between the underlying nodes and the fog layer on the premise of qualified user service experience and reduction of the system's service overhead,the thesis designs a low-coupling sensor-cloud based on double-buffer queue: command queue and data cache queue.The scheme puts the user's control information transmitted through the cloud in the command queue.The fog layer merges multiple repeated operation commands transmitted from the cloud,so as to the number of calls to the physical node,which improves the resource utilization of the node.The data cache queue is designed based on the KNN algorithm,which achieves fair cache based on factors such as priority,the remaining energy of the node and the number of hops to the fog node.Compared with the four algorithms of FIFO algorithm,SJF algorithm,EKM algorithm,and EKMB algorithm,the experimental results show that even if the experimental settings are different,the proposed EKMDB will reduce the number of rounds by 48.3% to 58.6% and increase the resource utilization by 52.3% ? 59.4%.(3)In order to effectively solve the problem of coupling task unloading in sensor-cloud,a dynamic unloading method of coupling task is proposed.Considering the restrictive factors of the fog layer,whose real-time state will determine the performance of data processing.In this thesis,a model is established,aiming to minimize the total processing time of offloading between the fog layer and the cloud.Since this problem is an NP problem,this thesis proposes a heuristic algorithm: first classify tasks,and different types of tasks are set different maximum tolerance delay time.Based on real-time status of the fog layer,the comparison of the completion time is calculated to determine whether the task should be offloaded to the fog or the cloud.Comparing to execute all the tasks by fog and cloud separately,the simulation results show that the proposed strategy can obtain the shortest total processing time and the minimum energy consumption.