We present an in-depth analysis of the crash-recovery problem and propose a novel approach to recover from otherwise fatal operating system (OS) crashes. We show how an unconventional, but careful, OS design, aided by automatic compiler-based code instrumentation, offers a practical solution towards the survivability of the entire system. Current results are encouraging and show that our approach is able to recover even the most critical OS subsystems without exposing the failure to user applications or hampering the scalability of the system.
"We Crashed, Now What?"
Cristiano Giuffrida, Lorenzo Cavallaro, and Andrew S. Tanenbaum
In the Proceedings of the 6th Workshop on Hot Topics in System Dependability (HotDep '10), October 3, 2010, Vancouver, BC, Canada