Parallel and high-performance computing

In a task-parallel runtime, some workers become idle while another worker still has a queue of ready tasks. Which scheduling approach is intended to help balance this kind of uneven workload?