Author:
Campos João R.,Nogueira Rodrigo Pato
Abstract
Software systems are used to execute critical tasks on a daily basis. Failures can easily lead to significant losses or even loss of lives. Online Failure Prediction (OFP) tries to predict incoming failures using the current state of the system. This relies on the premise that there are symptoms (i.e., some misbehavior of the system) prior to failure, however, characterizing the (mis)behavior of a complex system is an open issue. How can we know if the failure predictors are actually modeling the symptoms, and not just identifying correlations in the data? In this work, we explore the use of Statistical Process Control (SPC) to characterize the stability and instability of the Linux Operating System (OS).
Publisher
Sociedade Brasileira de Computação - SBC