ABSTRACT
This Dataspaces style applies to concurrent and distributed systems. It is a particular kind of shared-memory style: many independently executing information processing units consume data from a common substrate and produce data onto that, or other, substrate. This means that, very likely, different worker threads will be counting different occurrences of the same words, so each one produces only a partial word count. This style is particularly well suited for data-intensive parallelism, especially when the task scales horizontally, i.e. when the problem can be partitioned among an arbitrary number of processing units. The Dataspaces style was first formulated as such within the Linda programming language in the early 1980s. The model was put forward as a viable alternative to shared memory in parallel programming systems.
