Dynamic load distribution in massively parallel architectures: The parallel objects example