| In the distributed processing environment, there are many diverse approaches in studying a hypothesis. This is partially due to a lack of a standardized software product which could support distributed system applications. The design of the Distributed System Testbed is one attempt at reducing the need for such a software product. The testbed allows for distributed processing research and analysis by establishing the unique conditions necessary for the execution of distributed processing algorithms.; This project designed an architecture with three levels. The first level determines the elementary or low-level functions needed to support a testbed. Along with local resource management, these functions include simulating the behavior of a standard network type (routing and message-passing) on a physical network and creating a virtual configuration.; The second level identifies the higher-level functions associated with a distributed data environment. These functions manage the data topology of the entire network by simulating a virtual file from all existing attributes. Each user requests a subset of this virtual file or one logical record. Conceptually, this logical record can be described as a Virtual Distributed Data Structure (VDDS). A VDDS requires no physical interdependencies of any data fragments and is dynamically created.; The third level deals with the application-support functions (e.g. distributed database). These functions include transaction processing features and distributed concurrency control algorithms.; The algorithms supervise accessing local data fragments and can be interpreted as performing a Virtual Access Logic (VAL). Having VAL manage a VDDS realizes a Virtual Distributed Access Method (VDAM).; The evaluation aspect of the system focuses on module reactions to resource manipulations. This allows for greater freedom in transaction profiles and more complete statistics for recorded parameters.; The result of this project is a clearly-defined set of specifications which will encourage the implementation of this system. These specifications include the system architecture, functional description of all components (and subcomponents), transaction processing events (i.e. control flow and data paths) and evaluation measures. By providing this system, a standard for comparison of distributed processing research could exist which will benefit the computer science community.; All software is designed as modular, message-passing subsystems that facilitate extensions and modifications. This flexibility allows for the insertion of new modules and the replacement of corrected ones. |
