Brussels / 2 & 3 February 2013


Romain: OS Support for Replicating Binary Applications

Operating systems research in fault tolerance has mainly focussed on dealing with faulty software, while resilience against transient hardware errors has been dealt with mostly by applying compiler techniques or using RAD-hardened hardware components. L4Re's Romain replication framework allows to replicate binary user applications on top of the Fiasco.OC microkernel. Replicas are run independently without modifications and are validated whenever they perform externalization events, such as system calls. In the talk I'm going to introduce Romain's architecture and discuss issues we had to deal with regarding replication of multithreaded applications in a capability-based system.


Björn Döbel