Supporting consistent updates in replicated multidatabase systems |
| |
Authors: | Weimin Du PhD Ahmed K Elmagarmid PhD Won Kim PhD Omran Bukhres PhD |
| |
Affiliation: | (1) Hewlett-Packard Laboratories, 1501 Page Mill Road, 94304 Palo Alto, CA;(2) Computer Sciences Department, Purdue University, 47907 W. Lafayette, IN;(3) CEO, UniSQL, Inc., 9390 Research Blvd., 78759 Austin, TX |
| |
Abstract: | Replication is useful in multidatabase systems (MDBSs) because, as in traditional distributed database systems, it increases data availability in the presence of failures and decreases data retrieval costs by reading local or close copies of data. Concurrency control, however, is more difficult in replicated MDBSs than in ordinary distributed database systems. This is the case not only because local concurrency controllers may schedule global transactions inconsistently, but also because local transactions (at different sites) may access the same replicated data. In this article, we propose a decentralized concurrency control protocol for a replicated MDBS. The proposed strategy supports prompt and consistent updates of replicated data by both local and global applications without a central coordinator. |
| |
Keywords: | Multidatabases replicated data management concurrency control replica control serializability resolvable conflicts |
本文献已被 SpringerLink 等数据库收录! |
|