Bascinet: A kernel extension to tolerate hardware faults in user processes

Keun Soo Yim

Fault tolerance becomes more important problem in high-performance Linux systems (e.g., due to process technology of underlying hardware and complexity of software). This paper presents the design of an operating system kernel extension to detect hardware and software faults in user processes. System call requests generated by user process are continuously monitored to detect errors in the execution of the user process. A prototype system is implemented by extending a Linux-based platform. The performance overhead and fault detection coverage are evaluated.