Evaluating the efficiency of Linda implementations

摘要

Among the paradigms for parallel and distributed computing, the one popularized with Linda,rnand based on tuple spaces, is one of the least used, despite the fact of being intuitive, easy tornunderstand, and easy to use. A tuple space is a repository, where processes can add, withdraw,rnor read tuples by means of atomic operations. Tuplesmay contain different values, and processesrncan inspect their content via pattern matching. The lack of a reference implementation for thisrnparadigm has prevented its wide spreading. In this paper, first we perform an extensive analysisrnof a number of actual implementations of the tuple space paradigm and summarize their mainrnfeatures. Then, we select four such implementations and compare their performances on fourrndifferent case studies that aim at stressing different aspects of computing, such as communication,rndata manipulation, andCPU usage. After reasoning on strengths andweaknesses of the fourrnimplementations,we conclude with some recommendations for future work towards building anrneffective implementation of the tuple space paradigm.