Self-configuring Algorithm for Software Fault Tolerance in (n,k)-way Cluster Systems

Abstract

First of all, I would like to thank my advisor, Professor Shuki Bruck. His vision and guidance made this work happen. I would also like to express my gratitude to the other professors at Caltech, for you not only directly or indirectly helped my graduate studies, but also made Caltech a great place to be. The friends from the Paradise research lab and other labs helped make my Caltech years so enjoyable, that I don't mind being a graduate student for longer. My colleagues at Rainfinity made valuable contributions to some of the results in this dissertation. To them I say thank you. I dedicate this dissertation to my family  to my parents and my grandma, to my wife Fang and our precious little Hannah.

A Dependability Management Mechanism for Ubiquitous Computing Systems

“…The service time for a dynamic object is modeled according to hyper-exponential distribution. We performed experiments to find the optimal configuration for the ubiquitous computing systems using the following system-operating parameters [8,9,10]. The systems are analyzed at intervals of 1 year.…”

“…Software rejuvenation is based on the idea of preventive maintenance applied to the software context that is widely employed in different engineering disciplines [7,8,9,10]. Software rejuvenation is a proactive fault management technique aimed at cleaning up the system internal state to prevent the occurrence of more severe crash failures in the future.…”

“…So, in this paper, we propose a management mechanism for applying software rejuvenation technology into a ubiquitous computing system. It adopts the automatic monitoring scheme to collect, aggregate, filter, and track related metrics, the automatic analysis scheme to extend our previous work [10], the autonomic plan and execution scheme to suggest the optimal configuration alternative of a ubiquitous computing system. We validate the autonomic fault management scheme based on a workload model derived from the system log analysis.…”