Designing Distributed Computational Skeletons in D-Clean and D-Box

Functional programming has inherent parallel features. This situation can be exploited by building different language constructs for parallel and distributed programming. Earlier we designed two languages for skeleton based distributed functional programming on a cluster. The two languages have different expressive power and abstraction levels. Our distributed functional computation is based on a multiparadigm-oriented environment with several different layers. On the highest level D-Clean coordination language primitives are applied in order to define the distribution of the pure functional computation subtasks over a PC cluster. This distribution is made according to a predefined computational scheme, which is an algorithmic skeleton, parameterized by functions, types and data. The D-Clean programs are transformed into D-Box, a lower level description language including the computational nodes implemented in the lazy functional language Clean. The computational nodes are distributed over the cluster and they communicate using the middleware services. This paper presents skeleton based functional and distributed programming using the D-Clean and D-Box languages. The main goal is to illustrate the appropriateness and applicability of the previously introduced languages for distributed evaluation of the functional programs on clusters. Here, therefore, we design distributed functional computational skeletons and we have provided a set of known algorithmic skeletons implemented as D-Clean schemes and a set of D-Box programs.