Shared Data or Message-Passing - A Human Factor in Technical Choices?

Angela Sodan; Luiz Fernando Capretz

It is an ongoing debate in parallel processing whether shared- or distributed-memory computing models are better, whether shared-data or message passing is preferable. Recent research has shed some more light on the debate, showing that many applications can be supported well in either model (though potentially with some special tuning for the corresponding machine) but that for some applications with more extreme behavior the corresponding machine type and computing model are preferable or even the only feasible solution. Otherwise, what does make people choose one or the other? We investigate the human factor and propose the model that the personality type determines to a large extent personal preferences. The paper discusses the relationship between certain personality types and the programming model. To determine these aspects, we applied the psychological Myer-Briggs-Type-Indicator test on a group of students for whom both programming models were mostly new (i.e. who were not pre-occupied). The results give reasonable evidence for the validity of our proposed model and the relevance of the human factor in technical choices, i.e. that choices are not only/always a matter of which model is “better”.

Type of Publication: Paper
Conference: 15th International Conference on Parallel and Distributed Computing Systems
Publication Year: 2002
TitleShared Data or Message-Passing - A Human Factor in Technical Choices?
Publication TypePaper
AuthorsSodan, A, Capretz, LFernando