The Design Methodology of Distributed Computer Systems

摘要

This is a final report for research in distributed computer systems. Chapter 2 describes top-down development approach. The development process is divided into four successive phases: (1) requirement and specification phase; (2) design phase; (3) implementation phase; (4) evaluation and validation phase. Guidelines and automated tools for the first two phases are developed. A graphical method (using the max. flow min. cut algorithm and cut-tree concept) to decompose and partition a computer system into loosely coupled subsystems is proposed. The implementation and the evaluation and validation phases are outlined briefly only because they are very technology and architecture dependent. Chapter 3 examines Petri net model for asynchronous concurrent systems. Procedures based on Petri net for predicting and verifying the system performance are presented. The computational complexities of these procedures are also shown. Chapter 4 examines the analysis techniques for deadlocks in asynchronous concurrent systems. In particular, we study in detail deadlocks caused by conflicts in mutual exclusive access to resources with the constraint that each resource type has only one member. Chapter 5 first classifies and then evaluates several existing software reliability models according to some proposed criteria. Then it develops a theory of software reliability based on the nature of the input domain of the program, i.e., the size of the errors and the number, complexity and continuity of equivalence classes formed in the input domain. (Author)