Modern systems have extended their complexity, administration and design become almost impossible duties. The most frequent performance problems can be identified in database systems and more specific in distributed ones. The problems of data allocation and query optimization in distributed database systems are considered. Because of the possible wide distribution of such a system the replication problem in a distributed databases is presented by analogy with non-cooperative games. We consider the Santa Fe bar problem as an instance of the problem to automatically configure a replicated distributed database system so that its elements do not work at cross-purposes. We argue that a system designed according to this theory performs near to optimum and in particular avoids the tragedy of the commons for that problem achieving a global goal.