Random access protocols like ALOHA have been considered for machine-to-machine (M2M) communication in future networks for their simplicity of operation. This paper evaluates the performance of a Frame Slotted-ALOHA protocol that uses reservation and data packets (FSA-RDP), in a scenario where a controller collects data packets transmitted by a finite number of M2M devices. In FSA-RDP, frames of variable duration are divided in two parts, the reservation and data subframes. During the reservation subframe, active devices send short reservation packets to the controller. The controller assigns reserved slots in the data subframe to those devices that succeeded with the reservation. At devices, the FIFO service discipline and two queue management schemes, tail drop and push-out, have been considered. When the queue size is of one packet, we develop a discrete-time Markov chain to evaluate the protocol performance, including the cumulative distribution function of the delay of data packets that are successfully transmitted. Analytical results are validated by extensive simulations. The simulation model is also used to evaluate the system performance when larger queues are used. In addition, we study the impact that implementing Successive Interference Cancellation (SIC) at the controller has on the system performance. We also evaluate the performance of implementing SIC at the controller together with Irregular Repetition Slotted ALOHA (IRSA) to send the reservation packets. Numerical results show that the protocol efficiency of FSA-RDP is between one and two orders of magnitude larger than the efficiency of conventional Frame Slotted ALOHA, when a perfect channel is assumed. In more realistic channel environments, the use of SIC brings an important performance boost.