Research on Fault Tolerant Databases for Highly Parallel Real Time Distributed Systems.
Abstract
An open data inference technique is proposed which uses domain and summary knowledge to infer inaccessible data for query processing during network partitions. The open nature of data inference is due to the incomplete knowledge available about data and the need to combine partial inference results from separate processes to derive cooperative answers. To underlie such inference, new algebraic tools are developed for handling incomplete information. Further, a weaker correctness criterion, called toleration, is introduced to evaluate inference results. The above concepts have been implemented on a prototype Cooperative Distributed Database system, CDB, at UCLA. Our preliminary experimental results reveal that open inference can significantly improve the availability of distributed databases during network partitions.
Document Details
- Document Type
- Technical Report
- Publication Date
- Feb 28, 1992
- Accession Number
- ADA251627
Entities
People
- Wesley W. Chu
Organizations
- University of California, Los Angeles