Reinforcement Learning based on α-domination Strategy for Multi-criteria Decision Making and its Application to Distributed Database Systems