In the past, the performance of sequential programs grew exponentially as the performance of CPUs increased with Moore’s Law. Since 2005 however, performance improvements have come in the form of more parallel CPU cores. Writing parallel programs using existing programming languages can be difficult and error-prone. JStar is a new programming language that allows programs to be written in a naturally parallel way. The JStar project aims to produce compilers that can produce executables for a variety of architectures (such as many-core, GPUs and distributed computers). This thesis proposes a process for compiling these programs into distributed executables, and investigates various trade-offs and techniques for implementing JStar programs on a distributed computer. In this process, first a parallel design is created, then this design is expressed by a separate set of distribute statements that are combined with the original program to create a distributed program. The expressiveness and effectiveness of this approach is investigated for two case study JStar programs: (1) a prime number counting program (2) a version of Conway’s Game of Life. Various designs were hand-translated into a distributed Java programs and benchmarks were run to assess the performance of different designs. For each of these case studies, parallel designs were found that achieved high levels of speedup.