Research And Implementation Of Adaptive Query Processing Technology In Grid Database

The grid database is a new research content which combinds the technology of database and grid. It's a new demand put forward along with the development of grid, and its purpose is to integrate distributed heterogeneous databases into the grid environment effectively. Query processing technology is the core of the grid database study. However, the traditional query processing technology based on static optimization techniques can not adapt to the dynamic grid environment. Therefore, issues such as how to carry out adaptive query processing (AQP), adjust execution plan dynamically in the grid environment etc, have gradually become a hot area of the study domain.In this paper, we focus on techniques related to adaptive query processing in grid database. We propose an adaptive join query processing algorithm based on parallel pipeline, and then implement this algorithm in the grid database prototype system NHGridDB. The main contents and innovations of this thesis are as follows:⑴Introduce and analyze the grid database application environment and requirements. Propose the architecture of NHGridDB (a grid database prototype system based on OGSA), and then design a query processing architecture based on mobile agent technology.⑵Taking into account the dynamic characteristics of grid environment, propose an adaptive query processing framework. Present a cost-based model for pipelining processing of join query in grid database system, and then propose an adaptive join query processing algorithm AJAR based on this model. Algorithm AJAR incorporates techniques of random sampling etc. It can calculate costs of alternative execution paths in a dynamic fashion at runtime. These costs are then utilized to reorder the pipeline to adapt to changes in grid environment if a better alternative path is found.⑶Design and implement the query processing system of the grid database prototype system NHGridDB. Then implement algorithm AJAR and present experimental evaluation for algorithm AJAR with different data sets and parameters of grid environment. The results show that in most cases algorithm AJAR outperforms traditional query processing techniques.