Epsilon-consistent real-time monitoring for rapidly changing data

The problem of real-time monitoring in an open system is considered. In an open system, not only data would change rapidly but also the changes of the data might not be controllable. The notion of rapidly changing data as well as the consistency constraints in the domain of real-time monitoring is formally defined. Among all different types of constraints, only the problem of maintaining the $\epsilon$-Consistency constraints is addressed in this thesis. A real-time producers/consumers model, DONUT, is proposed to model this $\epsilon$-Consistent real-time monitoring problem. Then, a quantitative measure, resource freshness, is defined and used to indicate the performance of a real-time monitoring system. Several interesting problems related to the proposed real-time producers/consumers model are investigated: (1) average versus worst case freshness; (2) periodic producers/consumers with non-sharable resources; (3) non-sharable versus sharable resources; (4) aperiodic producers/consumers problems. We then discuss the applications of the DONUT model in the domains of network management and mobile networking. Finally, we show some experimental and simulation results.