In computing, dataflow is a broad concept, which has various meanings depending on the application and context. In the context of software architecture, data flow relates to stream processing or reactive programming.
Dataflow is a software paradigm based on the idea of disconnecting computational actors into stages (pipelines) that can execute concurrently. Dataflow can also be called stream processing or reactive programming.
There have been multiple data-flow/stream processing languages of various forms (see Stream processing). Data-flow hardware (see Dataflow architecture) is an alternative to the classic Von Neumann architecture. The most obvious example of data-flow programming is the subset known as reactive programming with spreadsheets. As a user enters new values, they are instantly transmitted to the next logical “actor” or formula for calculation.
Distributed data flows have also been proposed as a programming abstraction that captures the dynamics of distributed multi-protocols. The data-centric perspective characteristic of data flow programming promotes high-level functional specifications and simplifies formal reasoning about system components.
Hardware architectures for dataflow was a major topic in Computer architecture research in the 1970s and early 1980s. Jack Dennis of MIT pioneered the field of static dataflow architectures. Designs that use conventional memory addresses as data dependency tags are called static dataflow machines. These machines did not allow multiple instances of the same routines to be executed simultaneously because the simple tags could not differentiate between them. Designs that use Content-addressable memory are called dynamic dataflow machines by Arvind. They use tags in memory to facilitate parallelism. Data flows around the computer through the components of the computer. It gets entered from the input devices and can leave through output devices (printer etc.).
A dataflow network is a network of concurrently executing processes or automata that can communicate by sending data over channels (see message passing.)
In Kahn process networks, named after Gilles Kahn, the processes are determinate. This implies that each determinate process computes a continuous function from input streams to output streams, and that a network of determinate processes is itself determinate, thus computing a continuous function. This implies that the behavior of such networks can be described by a set of recursive equations, which can be solved using fixed point theory. The movement and transformation of the data is represented by a series of shapes and lines.
^ A Short Intro to Stream Processing