| Nowadays, the human society has been coming into a network-computing centered informationera, in which information sharing in great area and distributed transaction is becoming the urgent require of computer application. More and more applications want to access data in different network notes and different database. So multidatabase which can solve the interoperation of heterogeneous database problem is becoming a research hotspot in database area. Multidatabase can shield the user and application from the difference of local database, and provide global transaction management, and maintain consistency of data within the local database.Multidatabase technology has been accepted in the world. There are research organizations in lots of countries, and many interesting results have been presented. But because of the difficulty in resolving schematic heterogeneity, query processing, global transaction management and global view maintenance in multidatabase, there are still a lots of shortage in this research area. This paper researches in the key problems in multidatabase, and brings forward the algorithm to resolve these problems which remedy the shortage in multidatabase research area.To resolve the problem of multidatabase' complexity frame and difficulty realization, this paper brings forward object oriented multidatabase architecture based on CORBA and multidatabase middleware idea, which brings object oriented technology into multidatabase and makes the exploitation and use of multidatabase becoming more easy.Because of the characteristics of multidatabase, for example, local database in multidatabase has autonomy and transaction in LDB is unaware to global transaction manager, and so on, transaction management in multidatabase is very difficult. This paper analysis the shortage of traditional concurrency control method in multidatabase and brings froward a new concurrency control algorithm. We also bring forward a new recovery method for multidatabase: synthesis 2PC protocol, which combines many multidatabase recovery methods into one algorithm, and makes the algorithm more useful and more efficient.In multidatabase, access to global view will be decomposed to several sub-operations to several LDB, and perhaps also need data schema translation and query processing. If each accessing to global view must be translated into accessing to LDB, the cost will be very expensive. So there are materialized views in most multidatabase. The modification in LDB will cause the global view to be changed, so we have to do view maintenance in multidatabase with materialized views when the LDB changes. Most of the existing viewmaintenance algorithms only consider the views defined by select project join (SPJ) queries. Once the view is created by outjoin, those existing algorithms may fail in view maintenance. This paper introduce a new view maintenance algorithm, which can handle view involving outjoin, regular join and data inconsistency. Moreover, our maintenance mechanism minimizes the number of the queries that need to be sent to the local databasesystems for view maintenance, which is an efficient way to make the multidatabase system more efficient.Most of the traditional view maintenance algorithms only consider a single view, but there are many views in multidatabase, which contend a lot of information about LDB. If we make full use of these informations, to some views we can do view maintenance without accessing to LDB. This paper considers the problem of view self-maintenance, where the views are maintained without base database, including how to maintain a view and determining whether a view is maintainable.Deferred global view maintenance can speed up the update transactions by shifting maintenance into a deferred transaction, but even strict two phase locking(2PL) can not guarantee the serializable execution of transactions in this case. This paper develops a serializabitity theory base upon conflicts and serialization graphs in presence of deferred global view maintenance. This theory has...
|
PDF Full Text Request