The focus of this research is to develop a configurable client-side view storage system to reduce the cost of query processing and data shipping needs of distributed applications. The project aims to develop the theoretical basis of view management based on the notion of value order functions that imposes an order among a number of stored views. Value order functions can be defined by users employing a declarative language based on various properties of stored views. In this project, a prototype view management system is developed, called CAVES. It integrates view use and management functionalities into a single module with the help of two interconnected agents. The view use agent allows programmers to link view re-use methods to a database that helps CAVES maintain various statistics about the stored views. The view management agent applies general value ordering rules and optimization functions to this database to decide which views should be kept in storage. The effectiveness of different storage management policies and the performance of the system are tested with the help of a simulation model built on top of a high-performance parallel discrete-event simulation executive. The CAVES system will be used to improve the response time of client-server database systems by storing views that are computationally expensive or have high data shipping requirements at a client machine. The system will demonstrate that a dynamic collection of storage management protocols can outperform any static caching policy. The simulation system will provide the necessary methodology for selecting a set of protocols that are cost effective for a specific application. www.cs.rpi.edu/~sibel/research/CAVES/
