State transfer for hypervisor-based proactive recovery of heterogeneous replicated services
ISSN der Zeitschrift
Sicherheit 2010. Sicherheit, Schutz und Zuverlässigkeit
Regular Research Papers
Gesellschaft für Informatik e.V.
Intrusion-tolerant replication enables the construction of systems that tolerate a finite number of malicious faults. An arbitrary number of faults can be tolerated during system lifetime if faults are eliminated periodically by proactive recovery. The periodic rejuvenation of stateful replicas requires the transfer and validation of the replica state. This paper presents two novel efficient state transfer protocols for a hypervisorbased replication architecture that supports proactive recovery. Our approach handles heterogeneous replicas, and allows changing/updating the replica implementation on each recovery. We harness virtualization for an efficient state transfer between “old” and “new” replicas in virtual machines on the same physical host, and use copy-on-write disk snapshots for low-intrusive recovery of replicas in parallel with service execution. We apply the generic algorithm to a realistic three-tier application (RUBiS) and study the impact of recovery and state transfer on system performance.