Skip to main content
U.S. Department of Energy
Office of Scientific and Technical Information

Model of query processing and network support in a distributed database machine. [MUFFIN system]

Thesis/Dissertation ·
OSTI ID:5791003
Database machines are fundamentally distributed processing systems; however, they are based on several different approaches, associative disks, specialized CPUs, and conventional multiprocessor technology. It is not clear which of these approaches, or which particular designs, will be successful. This dissertation presents a step toward evaluating the performance of some of these database machines. An abstract model based on a distributed database machine, the MUFFIN system, was developed. The Simulation Program for the Analysis of Database Machines and Environments (SPADE), was then implemented to evaluate MUFFIN, but could be used to model other database machines. SPADE is comprised of two basic parts: the node model, and the network model. The node model was designed to simulate the activity of distributed database software. The network model was designed to support interprocess (and node-to-node) communications. The simulation experiments examined the performance implications of distributed query processing and network support. Results indicate that some performance improvement could be gained by using the fragmented query processing technique and by decreasing the cost of messages and disk I/O. However, the greatest improvement in system throughput was observed when the cost of page-level database operations was decreased.
Research Organization:
California Univ., Berkeley (USA)
OSTI ID:
5791003
Country of Publication:
United States
Language:
English