Energy Estimation And Optimizing The Energy-efficiency Of The Join Query Based On Cluster Greenplum Database

In modern information society, Massive data have been generated in every industry. Data become the basic of the industries as well as the important source of technolgy innovation. The foundation of the field of massive data is called data center, which store, manage and analyze the data. Both the small organizations owned several servers and the global companies regard data center as the necessary instrument to produce and develop. However, with the scaling of the data center and the background of global energy source problems, the energy consumption of data center is not only the key reason to constrain the develop of company leading by increasing cost, but also the new controdiction between the increase of economy and the waste of energy source.Database is the major application of the traditional data center. Since the great cost to maintenance data in the data center of large companies, research has focused on the energy comsumption of database system. In database systems, the query optimizer evaluates different computational paths by explicitly labeling their resource consumption. As the result, estimating the energy consumption of the query by query plan is the basic to research the energy of entire database system and also is the foundation to optimize the energy cost of query in database.Based on the clustered Greenplum database, this paper at first analyzes the difference between a single server database and the clustered database. Secondly, this paper researches the energy performance of the main energy cost hardware in different intense of workload and the energy cost of the query operators in different scale of tables. And thus according to the relation among hardware, query operators and intense of workload as well as the query plan, we build the model of energy estimation and experimentally examine the accuracy of the model. On the other hand, we present the strategies to optimize energy efficiency. We focus on three methods for operation of join: nestloop, mergejoin and hashjoin. We research on the computing time in different scale of operated tables among these methods. And finally this paper proposes the strategy to optimize the consumption of the problem of muti-join in database.