A novel framework for distributing computations DisPyTE – distributing Python tasks environment

Abstract

This paper proposes a novel framework for the distribution of concurrent tasks. DisPyTE (Distributing Python Tasks Environment) is written in the portable, interpreted programming language Python. It makes use of an event-driven, asynchronous network communication interface, which renders it especially well suited for application in heterogeneous network environments. After a short discussion on existing parallelization techniques, this paper illustrates the key principles of DisPyTE, including the main components and the call scheme. In a first example, it is demonstrated how DisPyTE can be used to distribute objective function evaluations in the scope of a heuristic optimization routine (genetic algorithm).