Riassunto analitico
The present work, written during a period of scientific collaboration with INFN
(National Institute of Nuclear Physics), proposes a mechanism for maintaining consistency
among replicated heterogeneous databases. This mechanism is integrated in a Replica Consistency Service for Data Grid environments.
Nowadays, an increasing number of applications, manage an enormous quantity of data distributed on a very large scale. In a Data Grid the data files are replicated to different sites. The replication of data among multiple databases is the only way to ensure that the data are available where and when they are needed.
The catalogs, used for maintaining information about the state of file replicas in a Data Grid, use replication to improve availability, to reduce data access latency, and to improve the performances of distributed applications. Moreover it is useful for fault tolerance.
A Data Grid may serve different user communities (Virtual Organizations), with different requirements. Therefore different types of data may be stored in different formats and in storage devices of different vendors. A mechanism is needed to maintain consistency among the replicas of a certain catalog that uses heterogeneous back-end databases.
In a Data Grid, heterogeneous data storage involves the problem of a uniform interface to access heterogeneous information; this problem requires tools for data integration.
Our goal for this work is to provide the ability to manage and to execute a replica synchronization; in particular we have a replica on an Oracle server and another replica on a MySQL server with the same logical schema. With the word "synchronization" we mean that an update of the target database will be done. Only
the changes that happen on the source database will be propagated to the target database.
|
