Database Systems as Middleware - Events, Notifications, Messages

Abstract

Database systems have been traditionally used as integration tools. Data integration was the primary goal of first generation systems. After relational technology had become a mature base for application development, new architectures where developed for the tight integration of data, procedures and all kinds of processing. This phase of DBS research and development was dominated by object-oriented database management and specific architectures like Starburst [HaCH 1990], which had a strong impact on current object-relational technology. The ubiquitous computer network has added another facet to the usage of DBS as an integration tool. In distributed environment, middleware aims at making distribution transparent. Corba or RMI are well-known examples. The call-oriented style of communication has been complemented by message-oriented, event-driven interaction of independent programs. System processes use this type of communication for decades. However, it is not well known as a mechanism on the application level - despite the fact that it has been employed for quite some time, e.g. in workflow systems [LeRo 2000]. The event-driven message passing paradigm becomes more and more important for highly distributed applications. Many kinds of interactions between applications follow a common pattern: n inde-pendent providers submit their output as messages, which in turn will be consumed asynchronously by m consumer applications. As opposed to call-level interaction, providers and consumers may or may not know each others identity. In a stock ticker application for example, there is no reason why providers should know the identity of consumers.