Slides of a presentation by Ted Dunning at BigData Everywhere conference in Boston on May 17th 2016.
"The leading edge of big data architectural practice is rapidly moving to flow-based computing using streaming architectures as opposed to state-based computing based on batch programs plus workflow schedulers. This transition is part of the larger movement towards micro-services and devOps oriented development and maintenance of large systems. This fashion is spreading quickly, but the understanding of why flow-based computing is different from state-based computing and what this means is practice is lagging behind. In fact, there is a huge difference and this has the potential of massively simplifying big data systems, thereby improving reliability and time to market. I will describe the necessary key concepts and illustrate them with practical examples. I will also describe why this matters in the real world.