Advanced Message Passing Infrastructure for Java

The present effort's purpose is to offer a first principles study of how to present MPI-like services to Java programs, in an upward compatible fashion. The purposes are twofold: performance and portability. For performance, we seek to take advantage of what has been learned since MPI-1 and MPI-2 were finalized, or which were ignored in MPI standardization for various reasons. The study will, for instance, draw on the body of knowledge just recently completed within the MPI/RT Forum, which strives to enhance both real-time and performance of message passing programs. From MPI/RT, we will at least glean design hints concerning channel abstractions, and the more direct use of object-oriented design for message passing than was done in MPI-1 or MPI-2 (despite existence of C++ bindings). Additionally, a fundamental look at data marshalling and unmarshalling in the Java context will be undertaken, and preference for Java-natural mechanisms and policies will be attempted. Along the lines of portability, a detachment from legacy implementations of Java over existing native methods will be emphasized, while also considering the possibility of layering the messaging middleware over standard transports and other Java-compliant middleware (such as CORBA). In a sense, the middleware developed at this level should offer a choice of a performance or generality emphasis, while always supporting portability. A policy/opportunity to support aspects of real-time and fault detection/fault-aware programs will be studied and standardized insofar as possible, again drawing on the concepts learned in the MPI/RT activity, and also drawing on experience from distributed computing real-time activities. The validity of this type of messaging middleware in the embedded and real-time Java application spaces will also be considered.