Low Cost Management of Replicated Data
Abstract
The aim of this thesis is to develop a method for managing replicated data that eliminates latency. Thus, our work can be viewed as a generalization of the results to systems with arbitrary types of data, with operations more powerful than reads and writes, and with other kinds of synchronization mechanisms than two-phase locking. Further, we show that it is possible to transmit the information required to maintain consistency using messages that would be sent between sites even if data were not replicated. In other words, the replication method requires no additional messages. The implication is that of the costs of replication described above, the only costs that need be incurred are the unavoidable costs of additional storage and local processing. A model for distributed systems is presented. Methods for managing replicated data is also presented.
Document Details
- Document Type
- Technical Report
- Publication Date
- Jan 01, 1986
- Accession Number
- ADA162771
Entities
People
- Thomas A. Joseph
Organizations
- Cornell University