The Model for Prediction Across Scales is a modeling framework for the rapid prototyping and development of dynamical cores intended for climate simulations. It is built upon an unstructured meshing framework that comes with its own unique performance challenges. Over the past few years, the development team for MPAS-Ocean (the ocean dynamical core within the MPAS framework) has been interacting closely with the SciDAC SUPER Institute to improve the performance of the model. Throughout the collaboration SUPER has provided assistance for general profiling, targeting areas of the model for improvement, and exploring alternative methods of decomposing the horizontal mesh. Planned future performance related advancements to MPAS include aggregated communications, custom non-blocking communications, multiple levels of parallelization (including threading), and advanced decompositions.