Design And Implementation Of Multi-task Concurrent Scheduling Strategy Recommendation System Based On Cloud Platforms

In recent years,the rapid development of large data,high-speed data processing capabilit y requirements are also getting higher and higher.While engineers are concerned about the actual business,the actual processing capabilit ies of the cloud platform are receiving more and more attention.At present,the world's major Internet C o have launched their own cloud computing platform for their own at the same time to provide external services.At the same time,the domestic government research institut ions are also building private cloud platform for dealing w ith high security data analys is business.With the continuous improvement of traffic,the various problems of cloud platform are gradually exposed,such as resource preemption,load balancing,task congestion and so on.Therefore,how to perform task scheduling and resource allocation is an important problem in the fie ld of cloud platform.In this paper,a method for generating and selecting mult i task concurrent scheduling polic ies based on task priority and task estimation execution time is proposed.A multi task concurrent scheduling polic y recommendation system for cloud platform is designed and implemented.The system can divide the homework of the PaaS cloud platform reasonably and select the optimal resource allocation for the operation of the atomic task,and select the partition of the node with the smallest execution time.In addit ion,aiming at the problem of high error in the prediction of execution time of task estimation,an error correction method based on decision tree is proposed to correct the prediction error.The design and Realization of the system for a single thread of tasks in different types of machine hardware CPU,memory card,and other factors under the task execution time distribution of mult i task scheduling strategy of cloud platform recommendation;time distribution to perform different tasks under resource ratio.The task segmentation method is used to segment the platform long task,so as to avoid the long-term task of using resources.Through mult i task scheduling strategy recommendation module is mainly through the time limit of task execution resources estimation task execution time based on the calculation of the ratio of the amount of resources and resources for implementation assessment of single atomic task.Through the analysis of the same type of task eva luation results performed using the amount of time and task execution resources,and according to the total execution time of the structure design of resources and tasks to predict job execution time of task execution to build risk assessment model,execut ion risk calculation task value.Multi task concurrent scheduling strategy is recommended by constructing a scheduling selection model analysis,and then through the ant colony scheduling algorithm and the minimum min scheduling algorithm,the task scheduling strategy is analyzed.The system will automatically select the optimal scheduling strategy for the task execution system or platform operators.In addition,the system uses machine learning method to learn the historical task execution time,the results optimization task scheduling strategy,the task execution time prediction module,and improve the prediction accuracy.Finally,this thesis tests the system of task scheduling,resource monitoring,mirroring configuration and monitoring,task concurrent scheduling and policy recommendation for the cloud platform multi task concurrent scheduling recommendation system.At the same time,according to the actual imp lementation of the system,the correctness of the proposed method is verified.In the end,all tests show that the system meets the design goals and performance requirements.