Research And Implementation On System-Wide Undo Recovery Techniques Oriented To Services

With the application of computer network technology more widely, various network services have become increasingly complex and fragile, and the failures related to network services have become commonplace. At present, human error has gone beyond software and hardware failure to become the most important factor which leads to system failures, therefore, how to solve the system failures caused by human error has lain ahead of people. Thus, system-wide undo techniques are presented to restore service system from failures caused by human error rapidly. System-wide undo mechanism provides operators a relaxed fault-tolerant environment and ability to retroactively repair, significantly reducing the difficulty and cost of failures recovery during the daily system management and maintenance, decreasing the MTTR, ultimately improving the dependability of service system.Firstly, an in-depth analysis of service-oriented system-wide undo mechanism is made, including basic concepts related to undo mechanism and design methodology of undo system, and on this basis a general undo system architecture based on 3R model is put forward, in which the key modules are designed and realized. Furthermore, an expansion of the basic framework for distributed services system is researched, with the improvement of interactive log structures and replay algorithm discussed preliminarily. Based on the conclusion of relevant researches a series of guidelines for undoable system design are presented, including undo-specific service design guidelines and general service design guidelines. Finally, an implementation method of undo system targeted at online auction service is discussed, and is verified by relevant tests.