Font Size: a A A

Object-oriented Databases Parallel Query Processing And Transaction Management,

Posted on:1999-12-11Degree:DoctorType:Dissertation
Country:ChinaCandidate:Y J WangFull Text:PDF
GTID:1118360182972347Subject:Computer Science and Technology
Abstract/Summary:PDF Full Text Request
The object-oriented databases(OODB)make up the drawbacks of the relational databases, and provide the powerful surport for many advanced database applications. However, the processing speed of associative query in OODB is slow, which influences the performance of OODB seriously and impedes the development and the application of OODB. There are several kinds of inherent parallelism in OODB, and the development of parallel processing technology becomes increasingly mature, so parallel query processing is an effective way to improve the performance of OODB. In this paper, according to the intrinsic characteristic of OODB and its query, based on the shared-nothing architecture, the stage-by-stage executing strategy ,the class-based hybrid data placement strategy,the mark-based data operation parallel executing algorithm ,the semi-join-based parallel query processing algorithm and relevant parallel query optimization strategies are proposed for query processing. Both theoretical analysis and simulation results show that they are better than traditional methods and capture the intrinsic characteristic of OODB and its query, and that they are practical and efficient. Cyclic query is a kind of particular and common query. In this paper, in view of the intrinsic characteristic of OODB and its cyclic query, according to the principle of "divide and rule", a practical and efficient cyclic query processing strategy is proposed, which is applicable to several kinds of cyclic queries. It has no limitation of the number of object classes, and can get the query results in polynimial time. Based on this, the parallel processing strategy for cyclic query is proposed. The efficiency of transaction management is critical to the performance of database systems. The nested transaction model is a complicated and flexible transaction model, and its application field is extensive. The nested transaction model had been listed in some international standard on OODB. The efficiency of nested transaction model is affected by several factors, the assignment of transaction identifiers and fault recovery are two important factors. In this paper, the requirements the transaction management pose on the transaction identifiers are analysed . Based on this , combined with the characteristic of nested transaction model, the bit-based transaction identifier assignment strategy, the improved transaction identifier assignment strategy and the optimal transaction identifier assignment strategy are proposed . The performance evaluation results compared with the traditional assignment strategy show that they satisfy the requirements of transaction management more efficiently, and that they are practical efficient assignment strategies. According to the characteristics of page server and nested transaction model in OODB, based on the Write Ahead Log protocol, the page-based recovery strategy——WAL_P is proposed, which is flexible and expansive, and can deal with all kinds of faults efficiently. In respect of implementation technology, the design and implementation of the client/server object-oriented database system—KDOODB is discussed .
Keywords/Search Tags:Object-oriented database, parallel query processing, data placement, parallel query optimization, cyclic query, transaction management, transaction identifier, recovery
PDF Full Text Request
Related items