Very Large Parallel Data Flow
Abstract
This report describes a three-phase effort to investigate the use of parallel processing in very large data/knowledge database management systems (D/ KBMS) as a means of attaining required performance levels. Phase one involved investigation of parallel processing techniques for inference processing, parallel processing techniques for very large database management and fault tolerant techniques for very large databases. During phase two, the results of ix investigation studies were combined to develop a methodology for specifying high performance, highly available D/KBMS computer architecture architectures for very large data/knowledge base (D/KB) environments. During phase three, a data/knowledge base management testbed was designed and implemented to serve both as a demonstration and performance measurement and evaluation platform. The testbed was used to perform several experiments designed to quantitatively measure D/KBMS performance and understand D/KBMS performance sensitivity and behavior with respect to various system paramounts. Several key results and conclusions have been identified from this work and directions for future work have been identified.
Document Details
- Document Type
- Technical Report
- Publication Date
- Mar 01, 1988
- Accession Number
- ADA196205
Entities
People
- Christine M. Baker
- H. Lu
- Jean L. Richardson
- K. Mikkilineni
- R. Ramnarayan
Organizations
- Honeywell International, Inc.