In this deliverable, we examine two practical scenarios in order to investigate how to overcome algorithmic limitations of the MapReduce paradigm to optimize execution speeds as well as the real-world applicability of preservation workflows defined in MapReduce. In particular, we give details on the optimization of an algorithmic operation that requires the use of iterations and give details on a large-scale preservation workflow in which we convert very large collections of images from TIFF to the JP2 file format in a distributed environment. We expand on the work reported in deliverable D6.2 in which we began formulating preservation workflows using the Apache Pig dataflow language as a higher order intermediary. Both the PPL translator and the large-scale preservation use case are now formulated in Apache Pig which in turn is then compiled down to MapReduce for execution in a distributed environment. Our findings indicate that preservation workflows can be formulated and executed efficiently within the boundaries of the MapReduce paradigm. Our PPL translator is available online at https://github.com/umaqsud/taverna-to-pig.
