This work shows the design and study of a family of algorithms that solves the multicast routing problem. In this problem, a given node called root has to send information to a certain group of receiving nodes. Although the algorithm can be applied at any level of the protocol stack, this paper studies its performance in the application level. This family of algorithms provides optimal routing tables between nodes belonging to the same multicast group, in such a way that the total transmission time is minimum. The algorithms take benefit from the delay time in the transmission of a message between one peer and another to forward the data to a third peer. Beginnig with a first algorithm, defined to send only one packet, some other algorithms has been described under certain conditions to send more than a packet with the maximum possible cadence and without congestion problems. With this purpose, we have restricted the number of times that the root may send a packet and also the maximum cadence time for the rest of the nodes. Moreover, we have applied mechanisms to guarantee full connectivity. With the aim of evaluating the performance of the different algorithms, we have calculated theoretically a set of bounds for transmission delays. Moreover, we present a serie of simulations over a virtual network that models an IP network. Over that first network, we have defined a second network of user nodes, which has been created at application level (so we can call it overlay network). We have applied the algorithms over the overlay networks, obtaining delay times, cadence times, number of nodes with congestion problems, and routing trees. Finally, we compare the results to check the best algorithm in any case. As expected, the fastest algorithms can usually have important congestion issues (more than a 50% of affected nodes). Moreover, the algorithm defined to avoid congestion has at most 50% bigger delay than the fastest algorithms, and hence we finally advice its application in multicast transmissions.