| The transformation of operators from traditional telecom service providers to digital service providers was accelerated by the development of Internet,communications and digital technology.In this process,value-added services in the field of telecommunications has been booming,more and more diversified telecom business,the user for mobile Internet traffic package,family number,call business cards and other value-added business orders generated by the rapid growth of orders.Order task granularity of different size,the length of the difference is huge,different types of business orders for the task of the machine's resource requirements are different,resulting in complex scheduling,prone to uneven allocation of tasks,low system resource utilization,high volatility and system operation Unstable problem.The main problems faced by the company's existing order task management system are as follows:Firstly,it is not ideal in controlling task flow and task allocation,and it is prone to more serious task throughput fluctuation,which makes the system prone to transient overload and even downtime.Secondly,when the system is designed to improve resource utilization,Task migration strategy,but the management of the task overhead is too high,taking up too much additional system resources,and sometimes to further reduce system utilization;In addition,the system uses a single coordinated server,not only prone to single point problems,reducing the system Of the reliability;and when the traffic is too large,the server will become a bottleneck of limited system traffic.In view of the shortcomings of the original order management system in task scheduling,this paper optimizes the task feedback control and proposes an active scheduling method which can realize the self-detection of resources.Then,we design and implement a new order with ZooKeeper Task management system in the task scheduling subsystem,the specific work includes:(1)In order to effectively control the order flow of tasks smoothly into the system.The PID control technology is introduced into the feedback control of the task flow,and the optimization strategy of integral separation and integral saturation threshold is adopted for PID control,the effect of feedback control is improved,the fluctuation of the system into the system is effectively restrained,and the stability of the system is improved.(2)In order to solve the problem of low utilization rate of existing system nodes,the task scheduling method of scheduling active transfer is designed and implemented,which can detect the resources themselves,reduce the management overhead and make full use of the resources of the nodes,At the same time to promote the node to do what,to ensure a healthy node work.(3)Based on the feedback control strategy and task scheduling method proposed in this paper,we design and implement a highly distributed distributed task scheduling system for telecom orders.At the same time,ZooKeeper is introduced to realize the characteristics of distributed coordination service.Solve the problem of single point of service and performance bottleneck in the original system,and deploy and test the system to verify the research results of this paper.In this paper,the distributed task scheduling system for telecom orders is a good solution to the problems of the existing order management system in dealing with node resources and the problem of entering the system task volume,and to avoid to a large extent.The effect of single point of failure on system reliability.The experimental results show that the CPU and memory utilization rate is 11.23%and 10.67%higher than that of the original system respectively.The average time of task completion is reduced by 9.52%,which suppresses the oscillations of the task flow into the system.The system is stable and reduces the economic cost.Operators get greater benefits,and for the sub-task scheduling scheduling initiative to provide new ideas to move. |
PDF Full Text Request