| Consistency has been a perpetual topic in distributed systems. The advent of peer-to-peer systems and sensor networks brings new challenges to the consistency problem. We recognize that given the nature of large scale and high dynamism of the new systems, probabilistic guarantee is a more natural replacement for the strict consistency guarantee pursued by traditional consistency algorithms. In this research, we proposes a number of techniques that facilitate probabilistic consistency for the new generation of distributed systems. We propose: a fast failure detection method, which is vital in improving the consistency of algorithms designed for a limited number of participants on top of p2p networks; a novel algorithm that selects servers according to user's criteria, which enables an efficient implementation of Probabilistic Quorum Systems (PQS), and hence all the applications that could be on top of PQS; an algorithm that achieves report data consistency while tolerates both false-negative and false-positive errors as the sensor deployment density grows. The new techniques will serve as the basis for new services to be built on top of the new generation distributed systems. We expect the scope of application to grow vastly with the consistency support enabled by the new techniques. |
