The Implementtation Of Data Accessing Encapsulation On Heterogeneous Data Source

Cloud computing is now one of the hottest technology in last few years. It is a distributed computing system which across over mass of computer. And just like gas, water and electric power, we can get it when we need it. Its potential commercial value has driving lots of enterprise to invest to it. But what difficulty is that accessing data based on heterogeneous data source structure. So we need to solve two problems in cloud computing system:1. provide uniform APIs(Application Program Interfaces) for application of upper layer;2. support transparent accessing of data on heterogeneous data source. These problems are also important for BI-PaaS(Business Intelligence and Platform as a Service) which is the project this research belong to.This paper is about data access layer module of BI-PaaS, and BI-PaaS is the subproject of one Research Institute Cloud Computing project. It is a BI information analysis platform based on cloud computing. And this research acts as data access layer of BI-PaaS, its most important target is struggling to resolve two problems above. Another target is we need to support HDFS(Hadoop Distribute File System) as one of the data source for most BI application.Through engineering practice, this paper researches the accessing of data on heterogeneous data source. Firstly, according to requirement of the project, analysed some main JAVA object persistent technology, and by contrasting these technology, we selected JDO(JAVA Data Object) as our technical plan. Then based on need of project, JDO specification and JDO product such as data nucleus, we designed our data accessing layer in detail.Then we figured out HDFS adapter based on achievement we gain above, made it possible for user to store JAVA object to HDFS by using JDO API. At last, do some viability test for HDFS adapter successfully.Finally, summarize all the work of this paper, and propose some existing problems and some optimizations for the next work.