An alternative to the sequential, or von Neumann–based, model of computing, offering an inherently parallel metaphor. The dataflow concept is characterized by an orientation toward producing and consuming data values rather than updating storage cells. Such systems are made up of dataflow machines which execute programs expressed as a partial ordering of operations. This partial ordering, called a dataflow graph, is created by high-level language compilers. Dataflow systems are capable of exploiting all of the parallelism inherent in an algorithm. The economies of scale of very large-scale integration can be exploited quite naturally by using the dataflow approach.