This study presents a newmethodology for the optimal allocation of switching devices in radial electrical distribution systems (EDSs). A specialised greedy randomised adaptive search procedure (GRASP) algorithm defines the location of a number of switching devices in order to simultaneously improve the following optimisation subproblems related to the use of the allocated switches: (i) the optimal reconfiguration of EDS and (ii) the optimal service restoration of EDS. Eventually, the objective function of the proposed switch allocation algorithm minimises the cost of the total expected energy not supplied, computed after deploying the service restoration, plus the cost of the total annual energy loss computed for every load level in a year, plus the investment costs associated with the number of installed switches. Both optimisation subproblems, i.e. the reconfiguration and the restoration of EDS, are represented by mixed-integer non-linear programming (MINLP) models and transformed into mixed-integer linear programming (MILP) models, using linearisation strategies. MILP models guarantee convergence to optimality by using convex optimisation techniques. Finally, all tests were carried out using a real 136-node distribution system, considering dispatchable and non-dispatchable distributed generation resources.