Providing fairness on shared-memory multiprocessors via process scheduling