Efficient aggregated deliveries with strong guarantees in event-based distributed systems
Abstract
A popular approach to designing large scale distributed systems is to follow an event-based approach. In an event-based approach, a set of software components interact by producing and consuming events. The event-based model allows for the decoupling of software components, allowing distributed systems to scale to a large number of components. Event correlation allows for higher order reasoning of events by constructing complex events from single, consumable events. In many cases, event correlation applications rely on centralized setups or broker overlay networks. In the case of centralized setups, the guarantees for complex event delivery are stronger, however, centralized setups create performance bottlenecks and single points of failure. With broker overlays, the performance and fault tolerance are improved but at the cost of weaker guarantees. The goal of this dissertation is to develop an efficient middleware for event correlation while still providing strong guarantees. First, we show what is necessary for strong guarantees in asynchronous distributed event-based systems that perform event correlation. Secondly, we provide the main deliverable of this dissertation: a generic middleware system, FAIDECS, which utilizes event types to efficiently correlate individually multicast events while providing strong guarantees for asynchronous event-based distributed systems. We then provide semantic alternatives to those provided in FAIDECS, showing what strong guarantees are able to be provided given certain operators.
Degree
Ph.D.
Advisors
Eugster, Purdue University.
Subject Area
Computer science
Off-Campus Purdue Users:
To access this dissertation, please log in to our
proxy server.