D Data Dissemination

One of the driving factors in IT development is the availability of cheap and efficient network technologies. The Internet is no longer used only as a medium for personal communication. Organizations start utilizing this technology to link existing applications and to create new ones. While traditionally systems were designed to respond to interactive user requests, they are now more and more aiming at autonomous, distributed data processing. Systems are connected and instantly react to changes to improve their functionality and utility (cf. zero latency enterprises). Mobile systems and other volatile configurations demand reactions to continuous changes; finance applications must be notified of price fluctuations; supply chain management must observe stock level changes; and information retrieval applications must forward new content (Banavar, Chandra, Strom, & Sturman, 1999; Gray, 2004). The focus when dealing with the delivery of data and services is changing, moving from a stationary world to one that is in a state of flux. Traditionally, data and services have been viewed as being stationary in a collection of objects or databases, with inquiries directed to them in a request/reply mode of interaction. This concept has led to client/server architectures that emphasize explicit delegation of functionality, where processes access remote functionality to accomplish their own goal. Remote procedure calls (RPC) and derivative techniques are classic examples (Birrell & Nelson, 1984; Mullender, 1993); even the incipient Web services mainly rely on sending requests with the Simple Object Access Protocol (SOAP; Alonso, Casati, Kuno, & Machiraju, 2003). These techniques deliberately draw from a successful history of engineering experience, their principles are well understood, and they have been an appropriate choice for many well-defined problems. In the context of dynamic or large-scale applications, however, request/reply has serious restrictions. The direct and often synchronous communication between clients and servers enforces a tight coupling of the communicating parties and impairs scalability (Franklin & Zdonik, 1998). Clients poll remote data sources, and they have to trade resource usage for data accuracy, especially in chains of dependent servers. Unnecessary requests due to short polling intervals waste resources, whereas long intervals increase update latency. The obvious need for asynchronous and decoupled operation has led to various extensions of existing middleware standards. For instance, CORBA and Java 2 Enterprise Edition (J2EE) were extended with asynchronous invocation methods and notification services (Object Management Group [OMG], 1999; Schmidt & Vinoski, 1999; Sun Microsystems, 2002). Database research, software engineering, and coordination theory corroborate the advantages of loosely coupled interaction (Cilia, Bornhövd, & Buchmann, 2001; Papadopoulos & Arbab, 1998; Sullivan & Notkin, 1992). The following presents a classification of communication paradigms and technologies to outline their fundamental characteristics.