Dataflow parallel programming language for clusters
PARUS is a data-flow parallel programming language that allows to build parallel programs for clusters and MPP multiprocessors. The data-flow graph is automatically converted to the C++/MPI source and linked with the libparus runtime library.
Also there are available tools for benchmarking cluster interconnect and visualize it.
Attempt of a portable cross-language object-oriented benchmark . So far supporting C++, Objective-C, and Java. Support for Smalltalk, CLOS, CSharp, and Eiffel is in development.