Font Size: a A A

The Research On Some Key Techniques Of Object Relational Data Base Management Systems

Posted on:2001-06-03Degree:DoctorType:Dissertation
Country:ChinaCandidate:G G YangFull Text:PDF
GTID:1118360092998897Subject:Computer Science and Technology
Abstract/Summary:PDF Full Text Request
Object-relational database technology combines the relational and object-oriented capabilities in one database system. It provides support for richer object structures and powerful query-oriented functions. This kind of DBMS can meet the requirement of some newly emerging application areas, such as CAx, GIS and OIS. Now, the Object-Relational DataBase (ORDB) technology is regarded as one of the most promising research and product development directions.This thesis studies the following key implementation techniques of ORDB, i.e., the query optimization techniques of ORDB, the join predicate processing and join predicate index techniques, the general index structure and concurrency control strategy, and the reference based join algorithms.In ORDB, user-defined time-consuming functions can be used in selection conditions, so the model of relational query optimization must be enhanced. Based on the System R style dynamic programming technique and the rank concept, a two-phase optimization algorithm Predicate_Rolling_Up is designed. Firstly, the System R style dynamic programming procedure (and the specific heuristic rule) is used to determine the join order of the relations. After that, the expensive predicates are placed in the operator tree (or execution plan) according their rank values with respect to joins. This method can guarantee the generation of optimization execution plan and costs a little.When user defined functions (UDFs) are involved in the join predicate, the algorithms for the computation of relational joins can not be used directly. In the paper, the restricting transformation function r, is defined, which can restrict some variable's values. Without scanning the entire relation, performance of join algorithm can be improved greatly. Generalizing the join index concept, a new detailed join algorithm based on predicate join index is presented, and the performance is analyzed.The index structure of an object-relational database must be extensible because its type system is. In the paper, an extended generalized search tree (EGiST) is presented to support the extensibility of data types and query predicates. A novel lock mechanism, which is a combination of predicate lock and two-phase locking of data records, is also designed in the context of EGiST. The Degree 3 consistency or repeatable read isolation of transactions can be guaranteed.Based on the reference feature of object-relational data model and the query language, the join algorithm Sort-Loops, Probe-Loops and Segment-Join are proposed, their performance are analyzed and compared. Under certain conditions, these algorithms are of good performance.In the end, there is a summary that contains a brief recap of each research chapter's main results and direction for future work.
Keywords/Search Tags:Query optimization, Index structure, Concurrency control, User-defined function, Query processing, Join algorithm, Algorithm analysis, ORDB.
PDF Full Text Request
Related items