Đang chuẩn bị liên kết để tải về tài liệu:
Error Log Processing for Accurate Failure Prediction

Predicting Needs as a Buyer. It is difficult to describe precisely the level of resources required to run an ex- periment or job. Depending on the inputs to a program, the ideal level of resource consumption can vary dramat- ically. Moreover, there is a tangible penalty formisestimating resource need, since these bids are made in advance of when the resources will actually be available. In order to match enough buyers with sellers, current market-based resource allocation schemes batch allocations into blocks of time. The time scale of this batch system can be min- utes or days ahead of when the resources will actually be made available. This means that users must predict their. | Error Log Processing for Accurate Failure Prediction Felix Salfner Steffen Tschirpke International Computer Science Institute Berkeley Humboldt-Universitat zu Berlin salfner@icsi.berkeley.edu tschirpk@informatik.hu-berlin.de Abstract Error logs are a fruitful source of information both for diagnosis as well as for proactive fault handling - however elaborate data preparation is necessary to filter out valuable pieces of information. In addition to the usage of well-known techniques we propose three algorithms a assignment of error IDs to error messages based on Lev-enshtein s edit distance b a clustering approach to group similar error sequences and c a statistical noise filtering algorithm. By experiments using data of a commercial telecommunication system we show that data preparation is an important step to achieve accurate error-based online failure prediction. 1 Introduction Despite of some early work such as 1 preparation of data has long been seen as the inevitable evil and has hence been neglected in most scientific papers. This applies especially to logfile data. However with the emergence of concepts such as IBM s autonomic computing 2 the importance of logfiles as a valuable source of information on a system s status continues to increase as can be seen from a variety of recent works such as 3 and the development of standards such as the Common Base Event 4 . This paper shows that clever mining of information from logfiles can significantly improve accuracy of an error-based online failure prediction method. However the goal is not to provide a comprehensive overview of various techniques that could be applied - we focus on a description of the techniques we have applied and how these techniques improved our results for online failure prediction. error events c B data window failure prediction present time Figure 1 Error-based online failure prediction. The term online failure prediction subsumes techniques that try to forecast the occurrence of system .