The Research And Implementation Of Dependability Evaluation And Test Tool Based On Event Injection

As computer systems are widely used in the dependability-critical fields such as aerospace, military and industry, the dependability of both software and hardware is becoming increasingly important. Moreover, computer networks make system failure more and more harmful, which expects more of dependable computing in the network environment. The event injection, a technique to accelerate system failure, can be applied to test and evaluate the performance of fault tolerant systems, which is very helpful to the implementation of highly dependable systems.In this paper the event classification and event modeling, event injection methods, concepts of dependability evaluation and test are firstly introduced. Then the development of dependability evaluation and test techniques are analyzed. Finally, the design and implementation of a dependability evaluation and test tool based event injection are presented.After study of the event injection theory and techniques used in the dependability evaluation and test, this paper designed the functional architecture of the tool, and then designed and implemented a system configure module, a heartbeat detection module, a management module, an event generation module, an event injection module, a result collection module and a data process module. Based on the analysis of the event classification and the properties of dependability, an event model is setup, which includes fault events, attack events and exception events. This paper also developed event injection technique for different event types and target systems, which support instantaneous, intermissive and permanent injections, and implemented concurrent injections as well.A fault-tolerance system is selected as the target system. According to the characteristics of the system, some events with different parameters are selected for injection. The result collection module collects data after the injection, and then these data are analyzed by the data process module. The effectiveness and feasibility of this tool are validated by computing the inject success rate, while the dependability of the target system is evaluated by computing the detect coverage and recover coverage.