Scalable Self Stabilization

The study of the study of self stabilization uses the theoretic model of some powerful adversary who can change the states of different parts of the system according to its wishes. This may look unrealistic (except in the context of hackers). However, It was shown that large distributed systems and network can enter arbitrary states because of a collection of faults, each of them being of a very simple nature (such as the loss of a message, or a crash). Hence, every such distributed system needs mechanism to self-stabilize, that is, to bring it from an arbitrary faulty state into a correct one. Initial self stabilizing protocols took global actions, involving the whole system, even for recovering from one isolated fault. In this talk we shall compare such earlier protocols to later ones that strive to scale better. For example, if only a single occurred, the recovery should be very fast.