New Fog Computing Architecture And Its Application In Remote Sensing Image Classification

As we all know,since the cloud computing was proposed in the late 1990 s,with the advantage of not sensitive to delay and network jitter in large-scale batch processing services,it has quickly become a commanding height of technology and a treasure land of wealth for all walks of life.However,with the rapid growth of the number of smart devices connected to the wireless network,the amount of edge data has reached the level of zettabytes(ZB),which has put tremendous pressure on the core network bandwidth of cloud computing;at the same time,the emergence of many emerging applications such as unmanned technology,location recognition,augmented reality,virtual reality etc,has put forward higher requirements on network delay,jitter,and data security.Traditional cloud computing has not been able to meet the above requirements well.People's demand for a high network environment and emerging applications like a fast response,low latency,and data security is increasingly urgent.The emergence of fog computing architecture might solve the above problems well.The fog computing architecture breaks the traditional centralized data processing method of cloud computing.It aims to process tasks submitted by users in the way of close to the user side.This way,the data does not need to be uploaded to the cloud computing center,which not only reduces the core network bandwidth but also greatly reduces the risk of data being stolen by uploaders when uploading.Because it is only one hop away from reaching the user,it is equivalent to processing locally for the user,which can provide users with low-latency and fastresponse services.The main work and research contents of this article are as follows:1.Based on the research on the leader election algorithm,Aiming at the problems of the current communication complexity of several leader election algorithms and the time-consuming leader election process,an improved leader election algorithm with dynamic serial number change is proposed.This algorithm numbers all fog computing nodes before the system starts.Based on the principle of small serial number priority,the fog computing node numbered 1 is directly appointed as the leader node,responsible for system resource allocation and management.When the leader node fails to exit,the node with the number 2 succeeds the leader node and becomes the leader node,meanwhile the current label of the node will be accordingly changed into 1,and then the rest of all nodes are notified to decrement their serial numbers by 1.When the node exiting the system rejoins the system,the serial number is automatically changed to n(10)1(n is the maximum value of the fog computing node number in the current system).Obviously,such a leader appointment mechanism can reduce the communication complexity of the algorithm.The algorithm communication complexity is reduced from 2O(n)to O(n),which shortens the execution time of the entire algorithm and improves the efficiency of the algorithm.2.Aiming at the problem that the task allocation in traditional fog computing architecture is not reasonable enough to make full use of system resources,a new fog computing architecture based on weighted rotation algorithm is proposed with the characteristics of fog computing architecture.First of all,according to the computing power and storage capacity of each fog computing node in the system,each fog computing node is assigned a corresponding weight value,the node with a higher weight value will be assigned to more tasks,likewise the node with a lower weight value will be assigned less tasks.In order to make full use of system resources.And prevent nodes from quitting due to faults and other tasks occupying computing resources at the same time,on the basis of the above mechanism,a mechanism for reporting the current task pre-completion time is introduced.Each node feeds back the current task pre-completion time in real time according to the task processing situation.The leader node updates the weight of each node in real time according to the collected information to make,the subsequent task allocation more reasonable,which naturally the purpose of improving the system resource utilization is achieved.3.This paper designs and implements an online remote sensing image classification system based on a new fog computing architecture through the research on fog computing architecture and algorithms.First,the definition,basic principles and classification process of remote sensing image classification are given.Then the system architecture,composition structure and workflow will be elaborated separately.Finally,the results of remote sensing image classification and rational analysis of the system prove the effectiveness and rationality of the proposed algorithm and the new fog computing architecture.The online remote sensing image classification system based on the new fog computing architecture can basically meet the user's needs for online remote sensing image classification business.According to the analysis of algorithmic communication complexity,the improved leader election algorithm proposed in this paper reduces the algorithmic communication complexity and can achieve the purpose of shortening the algorithm execution time.According to the comparison of simulation experiment results,the new fog computing architecture based on weighted rotation algorithm proposed in this paper is superior to the ordinary fog computing architecture.According to the analysis of the system operation results,the online remote sensing image classification system based on the new fog computing architecture can basically meet users' needs for the remote sensing image online classification system.