Reliable event messaging in big data enterprises
Tianning Zhang · 2015
Event-driven enterprise IT architectures are typically based on message broker systems. Most broker systems focus on functional decoupling, and on the fast and reliable delivery of events towards event consumers. Such consumer-centricity eases the development and deployment of consumers applications. However, especially within a web-based big data enterprise we typically have relatively small number of stable producers that is based on the core enterprise business model, and a higher, dynamic number of highly heterogeneous consumers. With the consumer-centric view of event messaging, problems with some consumers can easily lead to the congestion or service degradation of the whole messaging system, and can affect the operations of the producers or other critical consumers. To avoid such non-functional tight coupling, we propose in this paper a producer-centric architecture for messaging backbone in a big data enterprise. Within this architecture the broker infrastructure concentrates on the event production activities and pushes the delivery functionalities towards the consumer clients and applications. In this way we can achieve a higher level of SOA loose coupling, higher reliability and higher maintainability of the enterprise infrastructure.