In this section, some MAC power-saving protocols will be described. It has been observed in [3] that many power-saving schemes are based on the PSM to further improve the energy consumption under different circumstances. The study in [4] shows that the length of the ATIM window within the PSM significantly affects the routing performance. It is noted that the ATIM window is defined as the time interval for determining the MNs to be in the awake state for data transmission. The scheme in [5] modifies the 802.11 PSM by adjusting the length of the ATIM window based on the levels of network congestion. The Power Save Distributed Contention Control (PS-DCC) [6] determines its power-saving strategy by controlling the accesses to the shared transmission channel. The PAMAS protocol [7] utilizes busy tones in a separate signaling channel to control the packet transmission in the data channel. In additions, several routing layer schemes for power-saving purpose are proposed in [8] [9] [10]
[11]; while [12] presents a cross-layer power management system for the MANET. Power can be saved during the clustering stage [13] [14], and also by reducing contention in the MAC layer [15] [16]. Also research has been done based on IEEE 802.11s power management scheme, which puts idle stations (those with no traf- fic coming in or out) into sleep by shutting down their transceivers [17] [18] [19]. We will describe three representative MAC protocols; one is
asynchronous protocol; another is power controlling protocol; the other is topology controlling protocol.
2.3.1 Asynchronous MAC Protocols
Many MAC protocols are synchronous. In [20] there are three asynchronous power-saving protocols based on the IEEE 802.11 PS mode.
• Dominating-awake-interval protocol: The basic idea of this approach is to impose a PS node to stay awake sufficiently long so as to ensure that neighboring nodes can know each other. A PS node should stay awake for at least about half of BI in each beacon interval in dominating-awake-interval protocol. The sequence of beacon intervals is alternatively labeled as odd and even intervals. Odd and even intervals have different structures as defined below (see the illustration in Fig. 2.7). Each odd beacon interval starts with an active window. The active window is led by a beacon window and followed by an MTIM window. The function of MTIM window is like ATIM window. Each even beacon interval also starts with an active window, but the active window is terminated by an MTIM window followed by a beacon window. The above theory guarantees that a PS node is able to receive all its neighbors beacon frames in every two beacon intervals, if there is no collision in receiving the latters beacons. Since the response time for neighbor discovery is pretty short, this protocol is suitable for highly mobile environments. But, the power consumption is very large.
• Periodically-fully-awake-interval protocol: two types of beacon intervals are de-signed in this protocol to reduce the active time: low-power intervals and fully-awake intervals . Low-power intervals and fully-awake intervals have different structures as defined below (see the illustration in Fig. 2.8). Each low-power interval starts with an active window, which contains a beacon window followed by an MTIM window, such that Active Window(AW) = Beacon Window(BW) + MTIM Window(MW). In the rest of the time, the node can go to the sleep mode. Each fully-awake interval also starts with a beacon window followed by an MTIM window. However, the node must remain
Figure 2.7: Structures of Odd and Even Intervals in the Dominating-Awake-Interval Protocol
Figure 2.8: An Example of the Periodically-Fully-Awake-Interval Protocol with Fully-Awake Intervals Arrive Every p = 3 Beacon Intervals
awake in the rest of the time, i.e., AW = BI. The beacon intervals are classified as low-power and fully-awake intervals. The fully-awake intervals arrive periodically every p intervals, and the rest of the intervals are low-power intervals. If p is bigger, the power is more efficient.
• Quorum-based protocol: In this protocol, a PS node does not need to send beacons in every beacon interval. The sequence of beacon intervals is divided into sets starting from the first interval such that each continuous n2 beacon intervals are called a group, where n is a global parameter. In each group, the n2 intervals are arranged as a 2-dimensional n × n array in a row-major manner. On the n × n array, a node can arbitrarily pick one column and one row of entries and these 2n − 1 intervals are called
Figure 2.9: Examples of the Quorum-Based Protocol (a)Intersections of Two PS Nodes Quo-rum Intervals, (b)Node A’s QuoQuo-rum Intervals, and (c)Node B’s QuoQuo-rum Intervals
quorum intervals. The remaining n2− 2n + 1 intervals are called non-quorum intervals.
The structures of quorum and non-quorum intervals are formally defined below. Each quorum interval starts with a beacon window followed by an MTIM window. The node must remain awake for the rest of the interval, i.e., AW = BI. Each non-quorum interval starts with an MTIM window. After that, the node may go to the sleep mode, i.e., we let AW = MW. This is due to the fact that a column and a row in a matrix always have an intersection (see the illustration in Fig. 2.9(a)). Thus, two PS nodes may hear each other on the intersecting intervals. For example, in Fig. 2.9(b) and (c), node A selects intervals on row 0 and column 1 as its quorum intervals from a 4 × 4 matrix, while node B selects intervals on row 2 and column 2 as its quorum intervals. When perfectly synchronized, intervals 2 and 9 are the intersections.
Above three asynchronous MAC protocols are suitable for different environments. Dominating-awake-interval protocol is suitable for highly mobile environment but doesnt conserve energy since it requires PS node to keep active more than half of the BI. Periodically-fully-awake-interval protocol can save more power as long as p > 2 and is more appropriate for slowly mobile environments. Quorum-based protocol only need to send beacons in some BIs. The periodically-fully-awake-interval and the quorum-based protocols active ratios can be quite small as long as p and n, respectively are large enough. The Dominated-Awake-Interval
proto-col is most sensitive to neighbor changes, while the Quorum-based protoproto-col is least sensitive.
2.3.2 Power Controlling Protocols
Most MAC protocols are synchronous, and there are usually two broad ways to achieving less energy consumption. One is topology controlling, and another is power controlling. For example, The IEEE 802.11 PSM [2] is topology controlling protocol. The power controlling protocol is changing transmission power with packets.
The algorithm in [21] [22] determines different transmission and sensing ranges to optimize the transmission energy of the MNs.The algorithm belongs power controlling protocol. The node uses maximum power level to transmit RTS and CTS packets as illustrated in Fig. 3.5.
Then, the node calculate transmission power, which source node can transmit destination node suitably. Next, the node will transmit data packets with the calculated transmission power level. Because of lower transmitting power, the power consumption will decrease. The power controlling protocol also causes serious hidden node problem. As illustrated in Fig.
2.11, the transmission between node A and B will collide because of the transmission of node C and D. So, many protocols based on the idea of power controlling are developed. The study in [23] shows that the mobile nodes calculate the packet delivery cure, and modify transmission power. It shows that modes use adaptive power based on packet delivery curve. It is shown that CAPC [23] achieves similar throughput and higher throughput/energy consumption ratio than IEEE802.11 MAC protocol.
2.3.3 Topology Controlling Protocols
There are many topology controlling protocols. These protocols are usually based on the IEEE 802.11 PSM [2]. The algorithm in [24] has two functions. One is scheduling, and the other is dynamically adjusting the ATIM window size. The scheduling is proposed to avoid PS nodes contending medium again after the ATIM window without any extra overhead. Adjusting the ATIM window size dynamically is accommodating to various traffic conditions for improving network throughput and reducing PS nodes’ power consumption. The proposed protocol
Figure 2.10: The Power Control Scheme
Figure 2.11: Differences in Transmit Power Lead to Increased Collisions
in [24] has two main contributions. First, it avoids unnecessary frame collisions and backoff waiting time in data frame transmissions. Second, it conserve more power of PS nodes and improve to the channel utilization, and develop a intelligent strategy to dynamically shorten the ATIM window size. It is clear that reducing nodes backoff idle time and avoiding frame collisions can improve energy efficiency and network throughput. In this protocol [24] define that a buffered data frames duration, called working duration, is piggy-backed in an ATIM frame. To minimize average waiting time, it follow shortest job first policy basically, so the node with the shortest total working duration have the highest priority to transmit its buffering frames. The total working duration of a node is the sum of the working durations of all ATIM frames related it. The protocol [24] has special mechanism, which allows a PS node can go back to sleep when it completes all its data transmissions. Each node constructs the transmission table by the information in ATIM frames. Next, the nodes determine the transmission order. This is the scheduling transmission mechanism in the protocol [24]. There is no bound for the ATIM window size at the beginning of the beacon interval, and the end of the ATIM window depends on the traffic load of ATIM frames. Each node can obtain the duration of frame transmissions by overhearing ATIM frames, and calculate the total duration of all its currently receiving ATIM frames. If nodes sense the channel is idle more than TDIF S + TCW min , it deem that there will be no other node wanting to send ATIM frames. The ATIM window ends as illustrated in Fig. 2.12. If there is no any frame can be transmitted in the rest BI, all nodes end the current ATIM window immediately and enter the scheduled transmission as illustrated in Fig. 2.13.
Increasing spatial reuse is also a good way to reduce the energy consumption. As illustrated in Fig. 2.14, stations within the transmission range of A are all blocked due to the use of CSMA/CA mechanism when A is transmitting data to B with fixed power level. When A adjusts (decreases) its transmission power, some stations which is blocked by A will be released. Thus, C and D can transmit simultaneously when A is transmitting to B. Increasing spatial reuse is not only reducing power consumption but also increase the channel capture.
Above all, we know that the ATIM window size is relative to power consumption. When
Figure 2.12: The First ATIM Period Ending Rule
Figure 2.13: The Second ATIM Period Ending Rule
Figure 2.14: The Channel Capture Problem
the ATIM window size changes, the Beacon intervals should change. In this thesis, the analysis of beacon intervals will be discussed in next chapter.