Design And Implementation Of Multidimensional Analysis System In Digital Ocean

In order to welcome the coming of21century of marine, from the national development strategic, our country has put the "Digital Ocean" into the fifteenth and national marine development plan of2015, with the purpose of collecting marine three-dimensional monitoring information extensively, completing marine information engineering and the infrastructure construction of marine data. In order to manage and analyze mass of marine data, we must cluster marine data and build data warehouses. We are now building a multidimensional analysis system based on data warehouses, which can help us find some laws from mass of data in order to facilitate the decision-making related to marine scientific research persons.Currently there have been many of the world’s leading suppliers of data warehouse products, but these commercial products are expensive and not suitable for most enterprises, institutions or government agencies to use. Because the code is not open, it is not conducive to study. In response, the field of open source data warehouse project has been rapid development. There are some excellent open source products in ETL, OLAP and Data Mining. We design our multidimensional analysis system based on an open source OLAP engine Mondrian, the relational database ORACLE as data layer, Mondrian as OLAP engine, Jpivot as OLAP front-end show. Taking into account our proposal does not need EJB support, we use Tomcat as J2EE server. To extend our system, we have Mondrian’s source code in detail analysis and Jpivot’s source code in general. We also mainly focused on designing data model and building system architecture.We focused on a number of implementation details in multidimensional analysis system, including authorization management, schema file designing, some implementation details in multi-user environment and implementation of chart module. The multidimensional analysis system has been running and it is very well for marine science staff for decision support and reporting queries, which achieved the desired purpose.The data Warehouse is different from traditional transactional processing, which have the characteristics of analyzing large capacity data, so how to build a high-performance data warehouse system have become a hot topic of current research. Materialized view technology and indexing technology are two important technologies to enhance data warehouse performance. In order to improve multidimensional analysis system’s query efficiency, we presented a materialized view selection method based on workloads. The method first clustered queries in our logging system, then merging view technology was adopted within a cluster to generate candidate views, finally we would have materialized views with objective functions. We used these views to guide Mondrian’s query processing and improve the efficiency of query response. The results showed that it had greatly improved query response time and our method had good scalability. When storage space is limited, our method is still very effective.