Performance analysis of concurrency-control algorithms in distributed data-base systems

Thesis/Dissertation · Fri Jan 01 04:00:00 EST 1988

OSTI ID:5701957

Raghuram, A

Concurrent access to these data bases must be synchronized for proper execution of the user transactions and to maintain consistency of the data base. There are number of concurrency-control algorithms proposed in the literature for synchronizing the transactions, among which locking algorithms are the most popular. This thesis proposes analytical models for studying the performance of locking algorithms in distributed data-base systems. Models were developed for both fully replicated data bases and partitioned data bases. The models are based on mean-value analysis for data-contention and queueing-network models for resource contention. Only the average value of the data-base state, which is the number of locks locked, is used in modeling data contention. A simulator was developed for validating the analytical models and the results from the analytical models agree quite well with the simulation results. Results indicate that under certain restrictions, read-write access is equivalent to exclusive access and, also the same is true for non-uniform access of the data base. The network size has considerable effect on performance.

• Research Organization: Drexel Univ., Philadelphia, PA (USA)
• OSTI ID: 5701957
• Country of Publication: United States
• Language: English

Related Subjects

99 GENERAL AND MISCELLANEOUS
990210 -- Supercomputers-- (1987-1989)
990300* -- Information Handling
ALGORITHMS
COMPUTERIZED SIMULATION
DATA ANALYSIS
DATA BASE MANAGEMENT
DATA PROCESSING
DISTRIBUTED DATA PROCESSING
FUNCTIONAL MODELS
INFORMATION VALIDATION
MANAGEMENT
MATHEMATICAL LOGIC
PROCESSING
QUEUES
SIMULATION
TESTING
VALIDATION