Title :
Identifying network failure via detecting changes in power profile
Author :
Mahmoodi, Toktam ; Roveri, Manuel
Author_Institution :
Centre for Telecommun. Res., Kings Coll. London, London, UK
Abstract :
Failure in the network devices can be identified by detecting changes in the behaviour of the network. Such a network can be part of the communication infrastructure, which will play a significant role in the emergency managements. Network power consumption behaviour, namely its power profile, is a relatively robust profile such that from changes in that profile failure of the links or/and the nodes in the network can be detected. In this paper, we propose a statistical change detection system that relies on the Intersection of Confidence Intervals (ICI)-based test to detect variations in the profile of the power consumption. After failure detection, the proposed solution continues reading new samples of power consumption data to build the new power profile, which will allow us to detect multiple failures that occur in series. Through numerical examinations we show the effectiveness of the change detection action.
Keywords :
computer network reliability; power aware computing; statistical analysis; ICI-based test; emergency managements; intersection of confidence intervals-based test; link profile failure; multiple failure detection; network behaviour change detection; network device failure; network failure identification; network power consumption behaviour; power consumption profile variation detection; power profile change detection; statistical change detection system; Accuracy; Data models; Delay; Load modeling; Power demand; Stationary state; Training; Network failure detection; change detection test; network power model;
Conference_Titel :
Pervasive Computing and Communications Workshops (PERCOM Workshops), 2012 IEEE International Conference on
Conference_Location :
Lugano
Print_ISBN :
978-1-4673-0905-9
Electronic_ISBN :
978-1-4673-0906-6
DOI :
10.1109/PerComW.2012.6197614
