A Research On Low Delay Task Processing Technology With Edge And Cloud Collaboration Scenario

With the development of the Internet of things and the wide use of computers,various applications and services make people's quality of life more and more convenient.But at the same time,with explosive growth of the date,not only in the storage,a large number of cloud servers are occupied,especially the center of cloud computing equipment,it is easy to appear link congestion in the peak period.With the continuous improvement of people's quality of service and the continuous development of intelligent applications such as automatic driving,it is difficult for cloud computing to meet the requirement of the data transmission.At this time,putting part of the computing into edge computing devices has become a hot spot,which has been studied by operators,especially Internet of things providers.The research of task processing algorithm in the collaboration scenario of edge computing and cloud computing is particularly important.The purpose of this thesis is to study a task processing technology based on reinforcement learning in edge and cloud collaboration scenario.Limited by the hardware resources,this thesis mainly chooses to research on the simulation system designed by ourselves.In the case of ensuring the simulation effectiveness of the designed simulation system,as well as meeting the requirements of the real scene,the simulation system ensure that the algorithm designed in this thesis has good executability and can be applied in most of the edge and cloud collaboration scenarios under different hardware configurations in the real environment.Based on ifogsim,this thesis first proposes a system which can be used to predict bottlenecks and is suitable for resource scheduling simulation.The system inherits the core mechanism of event driven in ifogsim,decouples various components in the form of events,and greatly improves the scalability and modularity of the system.At the same time,in order to make the simulated results more accurate,the system can be used to simulate the bottleneck According to the real results,based on the idea of gradient descent and Markov process,this thesis realizes the function of automatic parameter adjustment and achieves good simulation results.On this basis,this thesis uses this system to design a more complex side cloud collaborative scene,and designs a task processing method suitable for the side cloud collaborative scene based on the policy gradient method in this scene,and achieves good scheduling effect.In summary,the innovations of this thesis are as follows.First,based on ifogsim,this thesis uses java to reproduce the core functions of ifogsim perfectly,and transplant it to Python environment,and provide a simulation platform for predicting bottlenecks and algorithm research.Secondly,on this system,an automatic parameter adjustment system is designed based on gradient descent algorithm,which is programmed by Java.Thirdly,this thesis proposes a reinforcement learning task processing algorithm based on the strategy gradient in the complex edge cloud collaborative scenario,and achieves good scheduling effect with low delay.