摘要
研究基于Linux网络服务器的软件老化过程建模、预测及软件再生策略问题.通过引入动量项及模拟退火算法的人工神经网络(BP)改进算法,对软件老化过程进行建模和预测.克服了BP算法收敛速度慢,易陷入局部极小值的缺陷,并提出将基于阈值的软件再生策略用于软件老化现象的主动性容错,提高软件系统的可靠性和可用性.仿真结果表明,该策略既降低了软件失效率,又减少了系统由于软件恢复而暂停服务的时间.
Software aging modeling prediction and software rejuvenation of Linux network server are studied. The momentum term and simulated annealing are added in the BP neural network to improve the convergence speed and overcome the local minimum value, by which the software aging is modeled and predicted. Meanwhile, a kind of software rejuvenation technique enhancing the reliability and usability of software system is put forward. Study results indicate that the method decreases the software disable rate and the unavailable time.
出处
《北京理工大学学报》
EI
CAS
CSCD
北大核心
2007年第7期625-629,共5页
Transactions of Beijing Institute of Technology
基金
国家自然科学基金资助项目(60472110)
关键词
软件老化
软件再生
BP神经网络
模拟退火算法
software aging
software rejuvenation
BP neural network
simulated annealing
