计算机工程与应用 ›› 2010, Vol. 46 ›› Issue (20): 126-128.DOI: 10.3778/j.issn.1002-8331.2010.20.036

• 人工智能 • 上一篇    下一篇

超级计算机错误预测模型研究

田曲波,邱德红,张奇峰,孙 蕾   

  1. 华中科技大学 软件学院,武汉 430074
  • 收稿日期:2010-04-14 修回日期:2010-05-14 出版日期:2010-07-11 发布日期:2010-07-11
  • 通讯作者: 田曲波

Analysis on failure prediction model of supercomputer

TIAN Qu-bo,QIU De-hong,ZHANG Qi-feng,SUN Lei   

  1. College of Software Engineering,Huazhong University of Science and Technology,Wuhan 430074,China
  • Received:2010-04-14 Revised:2010-05-14 Online:2010-07-11 Published:2010-07-11
  • Contact: TIAN Qu-bo

摘要: 错误预测对于提高计算机系统的运算稳定性有重要意义,日志分析是建立错误预测模型的有效方法。在同类型错误的时间预测模型的基础之上,通过日志分析建立了不同类型错误之间的关联模式,并在此基础上建立了基于关联模式的错误预测模型,填补了时间预测模型在错误发生后的短时间内无能为力的缺陷,提高了预测率,并在IBM的BlueGene/L的系统日志数据上验证了关联模式错误预测模型的有效性。

关键词: 关联模式, 错误预测, 日志分析, BlueGene/L

Abstract: Failure prediction is of great significance for improving the stability of the computer system and log analysis is an effective way of establishing the failure prediction model.On the basis of time prediction model for the same kind of failure,this article establishes the associated mode for different kinds of failure by the log analysis and the failure prediction model based on the associated mode,which fills the disfigurement of time prediction model’s powerlessness within a short time after the occurrence of failure and improves the rate of predication greatly.The validity of the failure prediction model based on associated mode is tested by the data of BlueGene/L’s system log of IBM.

Key words: associated mode, failure prediction, log analysis, BlueGene/L

中图分类号: