This thesis details ARMI, a parallel communication library that provides an advanced implementation of the remote method invocation protocol (RMI), which is well suited to object-oriented programs. ARMI is a framework for expressing fine-grain parallelism in C++, and mapping it to a particular machine using shared-memory and message passing library calls. It handles low-level details such as scheduling incoming communication and aggregating outgoing communication to coarsen parallelism when necessary. These details can be adapted for different platforms to allow user codes to achieve the highest possible performance without manual modification. ARMI is used by STAPL, a generic, object-oriented, parallel C++ library, to provide a portable communication infrastructure. The basic design decisions are described, as well as a detailed analysis of the mechanisms used in the current OpenMP/Pthreads, MPI, and mixed-mode implementations. ARMI's performance is compared with both hand-coded Pthreads and MPI on a variety of machines, including an HP-V2200, Origin 3800, IBM Regatta and IBM RS6000 SP.