The provision of services to the users of a multi-resource (10, 20, 30) communication network is controlled by modelling the system made up of these resources (10, 20, 30) as a Markov chain, wherein each state of the Markov chain is identified by a respective set of values of the numbers of the users served by each of said resources (10, 20, 30), and the transitions between states are representedby the allocation and de-allocation to the users of the services provided by the resources. A cost function is defined wherein each of the states gives a respective contribution weighted by the probability that the Markov chain is in that state, such probability being a function of the possible transitions between said states. A plurality of transitions between the states is thus identified that optimizes the cost function and the resources (10, 20, 30) are allocated the users according to such plurality of transitions that optimizes the cost function.