A lot of literature has been written about MCCA, its benefits, its limitations as well as how MCCA could be improved. We are going to briefly introduce the most relevant work regarding the improvement of MCCA, what their general idea, performance and limitations are ([15], [16], [17]).
4.1 eMCCA
In [17], the authors propose eMCCA (enhanced MCCA), an improvement of MCCA, where MCCA-enabled nodes can enjoy collision-free and guaranteed channel access during reserved periods. They achieve these features by creating mechanisms of prioritization and preemptive access. The prioritization is achieved by introducing a new interframe space, called MIFS (MCCA IFS) that is shorter than the interframe space values of non-MCCA nodes. That way, if the MCCAOP owner and a non-MCCA node compete with each other to access the medium, the MCCAOP owner will always win the right to access it, and non-MCCA node will find the channel busy after their interframe space periods. The preemptive access mechanism is achieved by restraining non-MCCA nodes from attempting transmission that could foreshorten a reserved MCCAOP. It will do so by preempting the channel access if the preemption time (the time between the end of the non-MCCA node transmission and the start of the MCCAOP) is lower than the estimated channel occupancy. In other words: if the channel is expected to be busy at the start of the MCCAOP. The MCCA node will start its MIFS period right at the end of the non-MCCA transmission preceding its MCCAOP. By using the preemptive mechanism, the MCCA owner actually accesses the channel earlier than planned, and is sure no ongoing transmission will shorten its MCCAOP.
According to the tests run by the authors, eMCCA does improve MCCA in term of network performance (the aggregate throughput is 25% higher), fairness among the different nodes (the proportion of channel access for non-MCCA and MCCA nodes is 60/40, instead of 90/10), collision rate with non-MCCA nodes (close to 0%), and percentage of time the channel is found busy (drop of 50%).
However the effectiveness of eMCCA is highly dependent on the estimations of the channel occupancy. An inaccurate estimation would make eMCCA a lot less efficient. Estimating the channel occupancy is something quite difficult to achieve, since the network conditions are often very variable and depend on a lot of different factors. The estimation mechanism would have to be highly dynamic and almost custom made for every network.
22
4.2 MARE
In [16], authors propose an alternative MAC protocol, called MARE (Medium Access through REservation) that is an alternative approach to reserve time slots for collision-free data transmission. It works on a similar basis as RTS/CTS handshake mechanism in 802.11 [18]. They propose the introduction of control frame for reservation request (RSV_REQ), reservation response (RSV_RSP) and reservation acknowledgement (RSV_ACK). RSV_REQ frames play a similar role to RTS in the way that when they are sent and heard, it indicates the medium is about to be reserved for transmission. RSV_RSP frames are similar to CTS frames in the sense that it confirms the RTS frame.
A node that wants to access the medium sends RSV_REQ to the intended receiver node, which will reply with RSV_RSP to agree or to disagree with the RSV_REQ. Upon disagreement, the RSV_RSP frame may include IE with parameters that would allow such a transmission (available bandwidth to transmit all the frames and forward them if needed). When the node receives the RSV_RSP frame, it will send in return a RSV_ACK frame to confirm the reception and the reservation of the medium.
Moreover MARE proposes to take into account the total fraction of the network bandwidth to be reserved by MARE-enabled node over non-MARE nodes, instead of a per-MP basis. Just like MCCA, MARE allows for contention-based data transmission during unreserved and/or unused time slots.
They also propose to accommodate the reservation, whenever possible, right after the DTIM beacon is transmitted, and to leave no time slots between reservation, so as to not let non-MARE stations access the medium.
The performance of MARE is better than MCCA, especially in the total overhead frames being transmitted. Since only the nodes having active reservations send control frames, the overhead traffic is less important (20% lower). MARE seems to be more effective than MCCA, regarding those aspects.
However, non-MCCA nodes may starve access to the channel. Indeed, by “forcing” reservations to start just after the DTIM beacon, and by aligning them without free time slots between them, non-MARE nodes may suffer from not being able to access the medium at all, for a long period of consecutive time slots. The resulting performance would be disastrous for those nodes, despite limiting the medium access for MARE nodes.
23
4.3 SMA
Finally, in [15], authors propose a enhanced mesh channel access method, called Scheduled Mesh Access (SMA), designed to address the issues of MCCA. SMA is quite similar to MCCA, in the sense that it also allows Mesh STAs to reserve time periods for transmission (in SMA, it is called SMA reservations). Yet, the channel access scheme is quite different: first of all, the SMA reservations used to transmit data are distributed across time among all the SMA-enabled Mesh STAs, in order to reduce self-interference. Second of all, if ever a Mesh STA is unable to transmit during its SMA reservation, it should continue accessing the channel even after the end of its SMA reservation, with only one exception: it cannot transmit during the SMA reservation advertised by other Mesh STAs. For instance, if a SMA-enabled node finds its SMA reservation to be shortened because of a busy channel at the beginning, it can continue transmitting its frames after the end of the reservation, as long as the channel is available, and that no other SMA-enabled Mesh STAs has a reservation scheduled next.
Third of all, in order to simplify implementation and ensure fairness among the SMA-enabled Mesh STAs, the contention parameters used to transmit data with SMA are based on access category of the frame and not the access mechanism used, therefore they are the same for SMA and non-SMA nodes.
The simulated performances of SMA are as follow: compared with MCCA, the packet loss rate drops significantly (with their network topology and parameters) to almost 0%, meaning all the packets that have to be transmitted are indeed transmitted. Moreover, the number of retransmission attempts per node also experiences a significant drop (from 3500 to less than 50), proving that SMA outperforms MCCA in this field. Since less retransmission means better network performance and less packet loss, we can expect SMA to perform better than MCCA in terms of throughput.
However, there are some limitations with SMA: indeed, as described earlier, SMA uses the same contention parameters for all nodes, it cannot ensure guaranteed access to the channel for SMA-enabled nodes. Moreover, there could be a scalability issue; since more SMA-nodes in the network would mean more SMA reservations and thus the channel would be less available for SMA-nodes to keep transmitting after their SMA reservation. The problem of shortened reservations would still not be solved.
24