Research On Native XML Database Transaction Management And Concurrency Control

With development of Internet, XML is being used in more and more areas as a new standard for data representation and exchange. However, how to store, query and analyze the increasing number of XML document become more and more important, with the increase of the amount of data available in XML, information management techniques for XML data become especially demanding. Today most relational, object-oriented or object-relational DBMS has provided extensions for XML. And at the same time, the database tailored for XML, namely XML native database, gains more and more attention of both researchers and producers. However, from a database point of view, all native XML databases today are far from perfect. Among all of the deficiencies, transaction management and concurrency control technique is one of the biggest ones. In this situation, it's of great importance that we investigate the transaction model and concurrency control technique for XML.First, on basis of traditional transaction management theory, we propose a general XML transaction model based on document tree and active tree instance, and define a set of XML document operations on the basis of this transaction model. We design a non-buffer logic and physical log based on extension and improvement of traditional recovery theory. In accordance with the characteristics of the structure of the XML document, we design a checkpoint and backup mechanism which can control synchronization between buffer areas of memory and data in disk. Considering XML concurrency control, we provide a concept of XML transaction serializability derive from traditional view serializability theory, and define conflict of XML operations and the detecting method of conflict.Finally, on basis of XML transaction mode, set of XML document operations and serializability theory, a locking protocol mechanism (XL) as well as a lock model is suggested, and the process regarding the application, release and update of the lock is illustrated in the paper. Besides, the schedule in this locking protocol mechanism is proved to be a serializable schedule.