Research on Software Systems that Keep Running Even If the Underlying Hardware Systems Stop

点击翻译按钮获取中文摘要

Even a brief stop of large scale online transaction systems will give critical damage to human daily life. The purpose of this research is to develop a mechanism that allows software systems to keep running even if their underlying hardware systems stop.In this research, I developed a middleware that allows objects in running Java programs to be cloned on multiple hardware systems. The middleware executes the clones of the objects synchronously. The entire systems can keep running, even if some of the cloned objects may stop, until all of the clones completely stop. The chance is quite low when all of the clones stop at the same time. I also applied the middleware to client-server systems to realize quite high available online transaction systems. I verified that the middleware works reasonably well to make existing systems more reliable. I also found that the approach using the middleware is quite useful to make the existing systems more reliable without modifying them.