Research On Microreboot Technologies Oriented To Self-Recovery

As mission-critical systems become more complex and heterogeneous, its security and availability become weaker, too. Academia focus on how to deduce MTTR because common methods improve MTTF to insure security and availability while improving MTTF becomes more difficult. Under this condition, microreboot which is a fast recovery technology comes into being. It not only heals a system by restarting failure parts, but also avoids data loss, process interrupt and long recovery time of whole reboot. Research on microreboot is little, and most concentrate in theory issues like constraints of microreboot, but few researchers focus on concrete microreboot methods. This paper researches on some aspects of microreboot as follows:Firstly, this paper analyzes the concept, basic principle and implementation strategies of microreboot in detail, then proposes an idea of crash-only software design, and summarizes the properties of crash-only software design.Secondly, this paper proposes a fine-grained microreboot method based on exception catching. After introduces the principle, work flow and key technologies of the method, this paper verifies the effectiveness of the method by experiment. The method realizes the recursion reboot of software systems, makes the self-recovery performance higher than the existing microreboot methods.Thirdly, this paper studies the automation of microreboot deep, and proposes an automatic execution method of microreboot based on Java. Then introduces the principle, flow and implementation of the method, and verifies the effectiveness by using an automatic processing software. The method improves the efficiency of the microreboot processing for Java systems, and ensures the safety and usability of them.At last, this paper researches the self-recovery technique of web application on the basis of existing achievement. Then proposes a self-recovery method of web application base on classified rebooting, and introduces the structure, flow and implementation of the classified rebooting. Finally this paper verifies the effectiveness by experiment. The method gives self-recovery abilities to Web systems, and ensures the safety and usability of Web systems.