Enhancing VLAM workflow model with MapReduce operations

MapReduce frameworks proved to be a good solution for storing and processing large amounts of data. Thanks to data parallelism, they allow to move computations very close to the storage and therefore to reduce an influence of "I/O bottleneck". Workflow Management Systems, in turn, are widely used for modeling of scientific applications. Users that are willing to use MapReduce frameworks in their workflows have to run separate environment to develop Map/Reduce operations. In this paper we propose an approach that will allow to extend existing application models by MapReduce routines. Our solution bases on DSL constructed on top of Ruby programming language. It follows examples of Sawzall and Pig Latin languages and allows to define Map/Reduce operations in minimalist way. Moreover, because the language is based on Ruby, the model allows to use user defined routines and existing Ruby libraries. A particular model of the workflow management system can be extended with our DSL letting users to use one environment for developing the workflow and MapReduce application.