Detail

Visigoth Fault Tolerance

Despite recent efforts to make the performance of data center networks and systems more predictable, replication protocols for stateful services have not taken advantage of such efforts; these protocols still make the worst-case assumption of an asynchronous system where all processes or messages can experience an arbitrarily long delay. The alternative assumption of a synchronous system is difficult to guarantee in practice and is not viable. Similarly, the principled approach for tolerating bit flips and data corruption is Byzantine fault tolerance (BFT). Unfortunately, BFT protocols make a worst-case assumption of an adversarial behavior, which is unlikely within the data center and increase the replication requirements. In practice, these services are designed optimistically assuming Crash fault tolerance (CFT) which is unable to capture arbitrary faults that surfaces at the data center scale.

In this talk, I will present a new technique for designing distributed protocols for building reliable stateful services called Visigoth Fault Tolerance (VFT). VFT introduces the Visigoth model, which makes it possible to calibrate the timing assumptions of a system using a threshold of slow processes or messages, and also to distinguish between non-malicious arbitrary faults and correlated attack scenarios. This enables solutions that leverage the characteristics of data center systems, namely their secure environment and predictable performance, in order to allow replicated systems to be more efficient with respect to the utilization of resources than those designed under asynchrony and Byzantine assumptions, while avoiding the need to make a system synchronous, or to restrict failure modes to silent crashes.