Research Of Scheduling Technique In Telecom Business Supporting Based On Cloud Computing Platform

Business Support System for Telecommunication holds up operation intergration of Telecommunication, information resources sharing , is a key to the success of every major telecommunication operator in current competitive circumstance. On the base of grid computing , Cloud computing is a new distributed computing technology and provides platform for solving large-scale calculation and mass storage. MapReduce is a common parallel programming and scheduling model used by the existing cloud computing system . MapReduce provides a strong and reliable parallel computing service and simplifies operation of administrators in cloud computing system. Task scheduling is an important component of MapReduce frame ,whereas ,the intrinsic task scheduling algorithm cannot adjust to the complex heterogeneous environment of Business Support System for Telecommunication in the MapReduce frame ,with a poor performance.This paper ,for Business Support System for Telecommunication, presents an improved Dynamic Threshold scheduling algorithm(called DT scheduling algorithm)based on the intrinsic task scheduling algorithm and the improved LATE scheduling algorithms. This algorithm uses real-time speed instead of mean speed when computing task remaining time .Real-time speed holds a higher timeliness and reliability comparing with mean speed ,therefore improves the accuracy of estimating task remaining time; Introducing the concept of Dynamic Threshold makes the decision threshold varies dynamically with change of task progress when judging backward tasks ,consequently it is possible to discover backward task earlier and more accurately .Define the weight of backup task when startup a backup task of backward task ,the executing order based on the weight in the task queue avoids startup unnecessary backup tasks ,saves system resources and raises the rate of computing resources .Finally ,by simulation experiments ,the effectiveness of DT scheduling algorithm has been verified.This paper is distinguished as : (1) using real-time speed instead of mean speed when computing task remaining time ; (2) using Dynamic Threshold to judge backward tasks, which varies dynamically with change of task progress ; (3) defining the weight of backup tasks of backward tasks and determining the executing order.