Network algebra (NA) is proposed as a uniform algebraic framework for the description (and analysis) of dataflow networks. The core of this algebraic setting is provided by an equational theory called Basic Network Algebra (BNA). It constitutes a selection of primitives and identities from the algebra of lownomials due to Stefanescu and Cazanescu.
Both synchronous and asynchronous dataflow networks are then investigated from the viewpoint of network algebra. To this end the NA primitives are defined such that the identities of BNA hold. These axioms are particularly strict about the role of the connections, which will be called flows of data. We describe three interpretations of the connections that satisfy the BNA identities: minimal stream delayers, stream delayers and stream retimers. Each of the above possibilities leads to a class of dataflow networks: synchronous dataflow networks, asynchronous dataflow networks and fully asynchronous dataflow networks, respectively.
For each case stream transformer and process algebra models are introduced and compared.
In this paper we prove the correctness of a specification of Synchronous Concurrent Algorithms in Process Algebra with process prefixing.