Studies on configuration and recovery techniques for fault-tolerant computing systems フォールトトレラントシステムに対する構成・回復技法に関する研究

博士論文

It is of great importance to operate a computer system with high reliability. Several techniques to achieve the high reliability of a computer system have been proposed and implemented in the real computer systems. This dissertation discusses configuration and recovery techniques for fault-tolerant computing systems, for which stochastic models are presented to evaluate performance and/or reliability. Chapter 1 gives introduction for configuration and recovery techniques based on the concept of redundancy. Chapter 2 presents two models for evaluating database recovery mechanisms. The first model discusses the recovery mechanism with periodical checkpoint generations. The second model further discusses the recovery mechanism in the situation where the road of the system varies with time in a shape of a cycle. Chapter 3 presents a model for evaluating the improvement on system reliability by retries based on time redundancy. In Chapter 4, two models for multi-processor systems are proposed from the viewpoint of transaction assignment, and are compared using the reliability/performance measures. Chapter 5 discusses a reliability evaluation software package tool for a system formulated by a continuous-time Markov chain with many states. Finally, Chapter 6 summarizes the results obtained in the dissertation, and discusses the further research works on configuration and recovery techniques.

Contents / p11 Abstract / p7 Acknowledgements / p9 Contents / p11 List of Figures / p15 List of Tables / p19 1 Introduction / p1 　1.1 Configuration and Recovery Techniques for Fault-Tolerance / p1 　1.2 Organization of Dissertation / p7 　References / p9 2 Evaluation for Database Recovery Mechanisms / p15 　2.1 A Recovery Mechanism with Periodical Checkpoint Generations / p15 　2.2 Checkpoint Generations in a Time Varying Load Situation / p37 　References / p55 3 Effects of Time Redundancy by Retries / p57 　3.1 Introduction / p57 　3.2 Model for a Retry Procedure / p60 　3.3 Analysis / p63 　3.4 Numerical Illustrations / p70 　3.5 Concluding Remarks / p82 　References / p86 4 Reliability/Performance Evaluation for Multi-Processor Systems from the Viewpoint of Transaction Assignments / p89 　4.1 Introduction / p89 　4.2 Multi-Processor System Modeling / p91 　4.3 Model 1 / p93 　4.4 Model 2 / p100 　4.5 Numerical Illustrations / p106 　4.6 Concluding Remarks / p111 　References / p114 5 A Reliability Evaluation Software Package Tool for Markov Models with Many States / p117 　5.1 Introduction / p117 　5.2 Mathematical Preliminaries / p119 　5.3 Applications / p126 　5.4 Concluding Remarks / p135 　References / p135 6 Conclusion / p139 Publications List of the Author / p145