Architectural alternative for data base

Thesis/Dissertation · Tue Jan 01 04:00:00 EST 1985

OSTI ID:5127458

Murphy, M C

The goal of this thesis is to determine which architectural features are viable alternatives for relational database machines. This requires consideration of both the algorithms used during query processing and of the underlying hardware used to support those algorithms. Two general approaches are proposed for augmenting database management system performance: specialized hardware (database machines) and fast access data structures coupled with query optimization. The database machine approach attempts to increase the performance of key functions by implementing them in either firmware or hardware. The query optimization approach attempts to minimize the computing resources required to execute a query by storing relations in specialized data structures and considering various alternative processing strategies. The purpose of this research is twofold: to study and describe relational database queries in sufficient detail to allow the functional resource requirements to be accurately ascertained; and to evaluate various uniprocessor and multiprocessor architectures in an implementation independent manner. Results include a new high performance algorithm for distributed relational join processing, a new definition for selectivity that is consistent and that permits unbiased estimation of query resource requirements, and detailed cost functions for both the uniprocessor and multiprocessor joint algorithms.

• Research Organization: California Univ., Berkeley (USA)
• OSTI ID: 5127458
• Country of Publication: United States
• Language: English

Related Subjects

99 GENERAL AND MISCELLANEOUS
990200* -- Mathematics & Computers
990300 -- Information Handling
ALGORITHMS
COMPUTER ARCHITECTURE
DATA PROCESSING
MATHEMATICAL LOGIC
OPTIMIZATION
PERFORMANCE
PROCESSING