Data-Oriented Architecture

2007 January " Data dominates. If you've chosen the right data structures and organized things well, the algorithms will almost always be self-evident. Data structures, not algorithms, are central to programming. Abstract As more devices and systems get woven into the fabric of our networked world, the scale and the complexity of integration is growing at a rapid pace. Our existing methodologies and training for system software design, rooted in principles of object-oriented design, that worked superbly for small scale systems begin to break down as we discover operational limits which requires frequent and unintended redesigns in programs year over year. Fundamentally, object-oriented thinking leads us to think in terms of tightly-coupled interactions that include strong state assumptions. Large scale distributed systems are often a mix of subsystems created by independent parties, often using different middleware technologies, with misaligned interfaces. Integrating such subsystems using object-oriented thinking poses some fundamental challenges: (1) it is brittle to incremental and independent development, where interfaces can change without notice; (2) there is often an "impedance mis-match" between subsystems in the quantity and the quality of information that must be exchanged between the two sides; (3) there is a real need to dynamically adapt in real-time to network topology reconfigurations and failures; (4) scalability, performance, and up-time cannot always be compromised in this dynamic environment. A different paradigm is needed in order to address these new challenges in a systematic manner. As the scale of the integration and complexity grows, the only unifying common denominators between disparate subsystems (generally numbering more than two) are: (1) the data they produce and consume; (2) the services they use and offer. In order to scale, system software architecture must be organized around a common "shared information model" that spans multiple systems. This leads us to the principle of "data-oriented" design: expose the data and hide the code. In this paper, we will discuss the principles of data-oriented thinking, and discuss why it offers an appropriate paradigm to address large scale system integration. We discuss the critical role played by the middleware infrastructure in applying data-oriented design, and describe a generic data-oriented integration architecture based on the data distribution service (DDS) middleware standard. We analyze popular architectural styles including data flow architecture, event driven architecture, and service oriented architecture from this perspective and establish that they can be viewed as specializations of the generic data-oriented architecture. Finally we illustrate …