Research On Cloud Computing Of BI Processing Technology

Business intelligence (BI) is a solution that combines with a variety of technologies and plays an important role in commerical information. But it is also has its limitations in use,including high cost, limited hardware resources,low security,high risk and so on. And its processing performance can not meet the functional requirements of extension.Therefore, only break through the limitations can develop business intelligence technology better.This article construct a new solution that combined cloud computing with BI technology to try to solve the limitations by analyzing the characteristics of cloud computing.First,select the Hadoop as the cloud computing platform for systemimplementation and experiment. Meanwhile we will improve the traditional cloudcomputing architectures and make cloud computing framework completelyservice-oriented from BI combination,called BIHadoop.Compared with the traditionalthree layer structure cloud computing,BIHadoop cluster structures adopted four layerstructure. At the top is a reverse proxy server, the second layer join the masternode.Super nodes play the essential role, not only make the nodes of two layersunder the direct control, but also can undertake the upper communications services. Itcan complete application service macro deployment, name management, server loadbalancing,front-end load balancing and risk control management and operationmaintenance inspection activities, and so on. The concrete structure of the first layer isagent layer;The second layer is the control node layer, consists of a super master node(main control node); third layer is the node name layer,and it is similar to the firstlayer of the traditional cloud computing architecture;The fourth floor is a super datalayer which is composed by many virtual machines. Compared with the traditionalHadoop cluster which has only one name node, this cluster has more than one namenode, the comprehensive performance of the system also has all aspects of upgrade.The nodes of the former two layers of the overall architecture can be classified ascontrol nodes. Each layer in the file system has unified management protocols, andmanages the special format of metadata to map with the nodes in the next layer at the same time.Next, we improve the BI architecture to adapt the Hadoop platform.First putforward the most important data mining module for the Hadoop platform improvedmodel in the BI system.Due to the traditional data mining system structure is gearedto the needs of single task processing serial structure.but,Cloud computing platformadopts the model of concurrent processing.So there is a bottleneck in data processingability and safety.Here,we build Hadoop platform oriented data mining cloudmodel.To improve the data mining model.There are HDFS data management,algorithm management and resource monitoring modules.The algorithm managementof the main module is based on many kinds of MapReduce algorithm integrationtoolbox(Data mining middleware),In order to better adapt to the cloud computingplatform of parallel computing and the MapReduce programming model.At the sametime makes data mining architecture combined with cloud computing technology ismore security and stability.And then to improve the BI system structure.We modulethe system structure according to the function needs.A total of eight classified asmajor function module.The improved architecture provides interfaces for the cloudcomputing to join in.Then we can get a better scalability and maintainability.Andform a powerful internal structure,scientific complete system structure with the cloudcomputing technology.We carried on the system implementaion after generating the cloud computing BItechnology solutions.and use the data mining algorithm to test the performance of dataprocessing.The test indicators are mainly load balancing features, storage capacity,costs, extension performance, safety performance and computational ability, the aboveindexes all have a better improve level.