Presentation 2006/7/25
Towards Fault Diagnosis for Large-Scale Distributed Systems
Naoya MARUYAMA, Satoshi MATSUOKA,
PDF Download Page PDF download Page Link
Abstract(in Japanese) (See Japanese page)
Abstract(in English) As distributed systems, such as clusters and grids, are getting larger scale and more commoditized, analysis of faults in such systems are becoming significantly harder than before. Nonetheless, none of the existing analysis techniques is not effective for such platforms, resulting huge burden to system administrators. We detect and analyze faults as follows. First, we take function-call traces from each process of the target distributed system. Next, to find anomalous behaviors, we apply an online analysis the call traces. Based on the premise that most of distributed systems processing is request-driven or event-driven, we analyze the call trace of each processing routine of requests or events. We implemented a prototype fault analyzer, applied it to a cluster resource manager, and evaluated the efficacy of our method.
Keyword(in Japanese) (See Japanese page)
Keyword(in English) fault analysis / data mining / distributed systems / clusters
Paper # DC2006-16
Date of Issue

Conference Information
Committee DC
Conference Date 2006/7/25(1days)
Place (in Japanese) (See Japanese page)
Place (in English)
Topics (in Japanese) (See Japanese page)
Topics (in English)
Chair
Vice Chair
Secretary
Assistant

Paper Information
Registration To Dependable Computing (DC)
Language JPN
Title (in Japanese) (See Japanese page)
Sub Title (in Japanese) (See Japanese page)
Title (in English) Towards Fault Diagnosis for Large-Scale Distributed Systems
Sub Title (in English)
Keyword(1) fault analysis
Keyword(2) data mining
Keyword(3) distributed systems
Keyword(4) clusters
1st Author's Name Naoya MARUYAMA
1st Author's Affiliation Tokyo Institute of Technology()
2nd Author's Name Satoshi MATSUOKA
2nd Author's Affiliation Tokyo Institute of Technology:National Institute of Informatics
Date 2006/7/25
Paper # DC2006-16
Volume (vol) vol.106
Number (no) 198
Page pp.pp.-
#Pages 6
Date of Issue