Chinese Journal of Computers   Full Text
Title	Research on the Application of Artificial Neural Network in the Fine-Grained Software Rejuvenation of Computing System
Authors	WANG Zhan GUO Cheng-Hao LIU Feng-Yu ZHANG Hong
Address	(College of Civil Aviation, Nanjing University of Aeronautics and Astronatics, Nanjing 210094)
Year	2008
Issue	No.7(1268—1275)
Abstract &   Background	Abstract In order to improve the software availability and reliability, intelligentize the software rejuvenation and boost up its veracity and efficiency, the rejuvenation granularity would be finer and artificial neural network would be applied. The key step of fine-grained software rejuvenation is to determine restart dependence between the modules. This paper researches the principium of artificial neural network, puts forward the model of artificial neural network which determines the degree of restart dependence between modules and reachable set of each module finally. Based on the coupling relation between modules of software system, this model analyzes the connection between restart dependence and coupling relation, sets down the arithmetic to calculate the degree of restart dependence between modules and reachable set of each module; so that the intelligent software rejuvenation with fine rejuvenation granularity is supported. Keywords software rejuvenation; rejuvenation granularity; degree of restart dependence; artificial neural network Background The Software Rejuvenation is a new technique of software fault tolerance which involves occasionally stopping the executing software, "cleaning" the "internal state" and restarting. This cleanup is done at desirable times during execution on a preventive basis so that unplanned failures, which result in higher costs compared to planned stopping, are avoid. So far, the detection of software fault as precondition of rejuvenation has been mature. The rejuvenation granularity has been fined to application layer, and the two-level-nested software rejuvenation policy has been researched. IBM has been pioneering software rejuvenation in conjunction with Duke University to improve reliability in both server and telecommunication environments. Now, the research on software rejuvenation focuses on there aspect: the thread-process-level rejuvenation policy, the intelligent rejuvenation policy, and the real -time rejuvenation policy. This paper puts forward the model of artificial neural network which determines the degree of restart dependence between modules and reachable set of each module, the model uses the calculation arithmetic of degree of restart dependence and reachable set by analysing the connection between restart dependence and coupling relation, so that the intelligent fine-grained thread-process layer software rejuvenation policy is supported. Then the software rejuvenation improves its veracity and efficiency, the software availability and reliability can be enhanced further. This work is supported by the Nation Natural Science Foundation of China under grants No.60273035. This project is conducted around the theories and principles in software performance maintenance. This research team has focus on the research of software rejuvenation for over four years. They have published over 15 papers in highly-ranked international conferences and journals in the field of software fault- detection mechanisms, the rejuvenation policy mechanisms, the architecture of software rejuvenation mechanisms, etc.