Load Balancing Strategies for Massively Parallel Architectures