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

Business Support System is the integration of telecom operators and the sharing of information resources support system. With the expansion of the information needs, the major operators adapting to changes in market competition, enhance the quality of service of telecom operators for centralized transformation to its business support system. Cloud computing as a next-generation distributed computing technology is a good platform for large-scale computing and mass storage. MapReduce is a software architecture proposed by Google, is mainly used for large-scale data sets of parallel computing. Task scheduling is an important part of the MapReduce framework, but the original task scheduling algorithm in MapReduce framework can not fully adapt to the complex telecommunications business support system for heterogeneous environments.Based on the characteristics of the telecom business support systems,there are some improvements to the DT (Dynamic Threshold scheduling algorithm improved dynamic threshold scheduling algorithm) algorithm: Define the node characteristics (static characteristics and dynamic characteristics), and present a task allocation algorithm based BP (error back propagation) neural network, it makes the task allocating better and improves the efficiency of the implementation of the entire operation; Improving the judgment conditions of the backward task, namely changing the formation conditions of the backup tasks make full use of the relative poor node and enhance the utilization of resources; before starting the backward task backup, all backups are not noly sorted according to a certain priority, but also based on the localization of data and node characteristics, thereby enhances the processing speed of the backup task, and reduces the working hours of the entire task. Finally, simulation results show the effectiveness of the above three improved algorithm.The innovation of the paper is: (1) the definition of node static and dynamic characteristics, and the introduction of the BP neural network algorithm to task allocation; (2) improving the backward task judgement algorithm; (3) starting backup task based on node characteristics and data localization factors.