Big Data Types : Internally Parallel in an Actor Language

Around year 2005 the hardware industry hit a power wall. It was no longer possible to drastically increasing computer performance through decreasing the transistors' size or increasing the clock-speed of the CPU. To ensure future development multi-core processors became the way to go. The Programming Languages Group at Uppsala University is developing a programming language called Encore that is developed to be scalable to future machines with a few hundred or even thousand processor cores. This thesis reports on the design and implementation of Big data types. Big data types are locally distributed data structures that allow internal parallelism in the actor model by using several actors in their implementations. Thus, rather than serializing all interaction these data structures are potentially as parallel as the number of actors used to construct them. The goal of Big data types is to provide a tool that makes it easier for an Encore programmer to create parallel and concurrent programs. As part of our evaluation, we have implemented a Mapreduce framework which showcase how of Big data types could be used in a more complex program