Distributed Bounded Predictability

Abstract

Complex systems often exhibit unexpected faults that are difficult to handle. A system that is predictable can anticipate faults and act accordingly. In particular, bounded predictability offers information that can be exploited by the system to adopt the best contingency plan. Such information is offered in form of a lower and upper bound. A lower bound guarantees a fault to occur after certain execution event, whereas an upper bound guarantees a fault to occur, in the future, but before some event. Complex systems are usually built from simpler components running concurrently. We study how to infer the bounded predictability property of a complex system (distributed and with multiple faults) from a parallel verification of bounded predictability of each of its components synchronising with fault free versions of the other one.