A Research Of Query Optimization In Embedded Mobile Real-time DBMS

Popularization of embedded devices, rapid development of mobile communication technology and wider using of real-time systems in the day-to-day work and life lead to increasing number of real-time mobile data which need to be managed on the embedded device. In this situation, the embedded real-time mobile database management system (EMRTDBMS), which combines the features of embedded system, mobile computing, real-time system as well as traditional database technology, has become a hot technology in the database fields, also the design of query optimizer for EMRTDBMS should been given full consideration to the changes of environment and make adjustments accordingly.The query optimizer of the database management system (DBMS) uses parse tree of the query as input, and then takes the steps of pre-processing, logic optimization and execution plan choosing, at last generates the final execution plan. During processing, heuristic rules and optimization technology based on cost model which depend on the data statistics in the data dictionary can be used.The logic optimization of the SPJ(Select-Project-Join) query mainly includes the processing of joining sequence choosing, conditions pulling down and projection pulling down. If there are"AND"or"OR"operation in the query conditions, it would effect the pulling down of the conditions. In the step of pre-processing, the query conditions can be transformed to normal form. After the comparison between the results of the logic optimization using the query conditions in conjunctive normal form and disjunctive normal, it can be said that the strategy of transforming the query conditions into conjunction form is better for the optimization afterwards.Embedded flash storage, data broadcast and the data deadline required by the real-time feature lead to the result that we can?t use traditional cost model in the ERTMDBMS. A new cost model considers all the factors which can affect the cost of the query in EMRTDBMS, including data source choosing, flash I/O cost, broadcast cost and data deadline, and an algorithm like dynamic programming which base on the new cost model can finish the processing of logic optimization and execution plan choosing.Through simulation analysis, the strategies of transforming the query condition into conjunctive normal form and the algorithm like dynamic programming for choosing the execution plan based on the new cost model can work well when the number of the relations is relative small.