Description of Handshaking Based Protocols

Karn [127] brought withMACAa substantial advance by introducing the Request-to-Send / Clear-Request-to-Send (RTS/CTS) handshaking for wireless networks (figure 1.11).

The idea of carrier sensing was totally eliminated due to the poor performance of CSMA in multi-hop networks. Hence, RTS is sent by S according to the ALOHA protocol, stabilized by a binary exponential back-off (BEB). If D correctly receives the RTS, it sends back a CTS packet. RTS and CTS include information on transmit packet data length. Any station hearing the CTS prevents itself from transmitting, so that the hidden terminal problem is partly solved. A station hearing the RTS but not the CTS is allowed to transmit, thus the exposed terminal issue is also reduced.

Moreover, only short control packets RTS or CTS are lost in case of collision: The time to solve the competition is reduced and the data packet doesn’t need to be retransmitted.

BEB may result in an unfair sharing of resources [54]. An improved version called MACAW suggests three modifications of the back-off algorithm used by MACA.

• The first one is the Multiplicative Increase and Linear Decrease (MILD) mech-anism: Upon a collision, the back-off window is multiplied by 3/2 and de-creased by one unit at each success.

• Data packets include a field in their header which contains the current value of the back-off window. Whenever a station hears a packet, it copies this value

S RTS

CTS

DATA t

t

RTS CTS

D

Nodes inhibited by the CTS Nodes inhibited

by the RTS Destination (D) Source (S)

Figure 1.11: MACA data transmission with RTS/CTS handshake.

into its own window size.

• The back-off algorithm is run in each station independently for each stream.

Finally, the MACA handshake is modified in a RTS / CTS / DS / DATA / ACK sequence. Before sending data, a short Data Sending packet is transmitted by the sender to inform that the RTS/CTS dialogue was successful. A link level acknowl-edgment, optional with MACA, is considered here as mandatory.

Adapting the handshaking to the family of busy-tone protocols, the Dual-BTMA [106] (DBTMA) improves the short dialogue with two busy-tone channels, one for the sender, the other for the receiver. This has the main aim to combat both hidden and exposed terminal problems. An extension of DBTMA for terminals with directional antennas (DBTMA/DA) is proposed in [114]. A comparison of protocols using this technique can be found in [113].

Again in the family of out-of-band signaling protocols, the Power Aware Multi-Access Protocol with Signaling (PAMAS) considers reducing the energy consump-tion at each node. In order to avoid consuming any power for receiving packets with destination address of other nodes, [175] suggests the use of the RTS/CTS handshake on a separate channel. This helps nodes to know when and for how long

they can power themselves off. For a survey on energy efficient MAC protocols for ad hoc networks see [141].

t

propagation time turn around time

the RTS sender

senses the channel busy t

Figure 1.12: In FAMA, the dominating CTS plays the role of a busy tone.

Since using two separate channels is more cumbersome, modifications have been performed, e.g. [95] proposes to use the CTS as a kind of in-band busy-tone. In FAMA protocols, CTS and RTS lengths are calculated with due consideration of the radio propagation delay, the processing time, and the turn-around time, i.e., the time for radio device to switch between transmission and reception states. Hence, a station that transmits a RTS simultaneously with the CTS of a receiver would hear at least a portion of the CTS at the end of its transmission. This is shown in figure 1.12. FAMA-NCS is based on carrier sensing, while FAMA-NPS is based on packet sensing.

Figure 1.13: Data transmission with the receiver-oriented MACA-BI protocol.

Consideration of turn-around time imposes a heavy penalty on the use of re-sources. It is further observed in [186] that the relevant area of contention is the reception range of the receiver. In the MACA By Invitation (MACA-BI) protocol, the sender waits for an invitationby the receiver in the form of a Ready-to-Receive (RTR) packet (figure 1.13). This simple handshake improves the performance of the MAC layer with respect to CSMA, FAMA, and MACA by reducing the overheads.

An additional advantage is the reduced power consumption. However, MACA-BI relies on traffic prediction at the receiver. It must indeed be able to approximately know when the sender has a ready packet. This can be helped by the indication in data packets of the buffer status at the sender side. The receiver-oriented notion is exploited in chapter 3 by CROMA.

The notion of quality of service (QoS) in the MACA family of protocols is in-troduced in [140] with MACA with Piggyback Reservations (MACA/PR). Non real-time packets are transmitted according to the RTS / CTS / DATA / ACK sequence. The first packet of a real-time flow is sent using the RTS/CTS handshake and reserves time intervals for subsequent packets along the flow path in the net-work. Then, each real-time packet and associated ACK piggybacks a reservation for the next packets. MACA/PR is coupled with a QoS routing algorithm.

DFWMAC: the Distributed Coordination Function of IEEE 802.11 The Distributed Foundation Wireless MAC (DFWMAC) is the protocol adopted by the IEEE 802.11 standard [35] for its so called Distributed Coordination Function (DCF). This scheme is the direct heir of MACA/MACAW on the one hand, and of CSMA on the other. From MACA, it has taken the handshake, the virtual carrier sensing, the BEB ; from MACAW, the ACK control packet ; from CSMA, the physical carrier sensing.

difs DATA

ACK Source

Destination Other

sifs

t t

contention window decremented

Figure 1.14: Basic mode access of IEEE 802.11 DCF.

DCF has two modes that can be dynamically chosen according to the transmit data packet length. In the basic mode (figure 1.14) rather used for short packets, a station ready to send picks a random number of slots (of length SlotTime) in its back-off window [0;CW]. The back-off timer is decremented at each time-slot only if the channel is sensed idle for more than a DCF inter-frame space (DIFS) interval.

The station is allowed to transmit when the timer reaches zero.

The contention window parameter (CW) has an initial valueCWmin. After each failure of the transmission, CW is doubled. It is however upper bounded byCWmax. CW is reset to the minimal value after each successful transmission.

Each correctly received frame is acknowledged by an ACK control packet. The interval between data and ACK is set to the Short inter-frame space (SIFS). Note that SIFS is shorter than DIFS, so that a receiver acknowledging a frame has priority over a new transmission.

Figure 1.15: RTS/CTS mode access and NAV setting in IEEE 802.11 DCF.

With this handshake procedure, DFWMAC takes into account the hidden ter-minal problem. The basic access described above is applied to the RTS transmission (figure 1.15). The short dialogue is not recommended for data packet lengths smaller than the RTS because in this case a collision on the data packet is less time con-suming.

In addition, a virtual carrier sensing is implemented. RTS, CTS and data pack-ets include the information related to the sequence duration. Stations hearing this field set their Network Allocation Vector (NAV) to one, indicating that the channel is busy for the time of the transfer. Finally, the protocol allows the successive trans-missions of several segments, each one individually acknowledged by the receiver.

Bianchi in [55] provides an accurate analysis of the saturation throughput of DCF in a fully connected network and for a finite number of terminals under ideal channel conditions. [60] analytically derives the average size of the contention window that maximizes the throughput and proposes to tune the back-off algorithm to increase the capacity. [68] studies the influence of hidden nodes on the performance of the two access modes of DCF.

Sometimes, IEEE 802.11 DCF has been criticized for lack in performance in multi-hop networks [202]. It is claimed that it has still the hidden terminal problem, it doesn’t solve the exposed terminal problem, and the back-off algorithm is judged to

cause unfairness, especially with TCP. Besides, questions regarding the effectiveness of the RTS/CTS handshake have been raised in [201]. If the interference range of radios is much larger than the transmission range, the efficiency of virtual carrier sensing is reduced.