Chapter 2 Overview of IEEE 802.11 Power Saving Mode and Related Works
2.2 Related Works
2.2.3 An energy efficient MAC protocol for IEEE 802.11 WLANs
In [9], Wu et al. proposed an energy efficiency MAC protocol for IEEE 802.11 networks
by scheduling transmission after the ATIM window and adjusting the ATIM window dynamically to adapt to the traffic Loads. But they did not consider the energy consumption caused by overhearing among neighboring nodes. Nodes schedule those to-be-transmitted data frames after ATIM window. According to a buffered data frame’s duration, nodes determine the transmission order. Data transmission takes place to avoid unnecessary frame collision and backoff time. Besides, nodes adjust the ATIM window dynamically to adapt to the traffic loads.
Chapter 3
Power-Saving MAC Protocol 3.1 Overview
The proposed power-saving MAC is an energy-efficient dynamically-adjustable ATIM window protocol designed for real- and non-real time data broadcasting. Under the proposed protocol, data is transmitted according to a dynamically updated transmission schedule. Each node utilizes prioritized contention based on Enhanced Distributed Channel Access (EDCA) for periodical real-time (CBR Voice) and non-real time (Data) traffic. Each node with real-time (CBR Voice) traffic initializes access to medium is through contention, but once a node reserves transmission time, its reservation time in the subsequent Beacon Interval continues automatically as long as the node continues to broadcast a packet in each Beacon Interval. Thus, nodes need only contend for transmission time at the beginning of Real-time (CBR Voice) traffic bursts in the ATIM window.
Nodes transmitting Real-time (CBR Voice) traffic also require the Reservation Occupation Table maintenance scheme to ensure correct channel utilization. The CBR Voice nodes transmit Beacon messages which carry the Reservation Occupation Table to each network node at the beginning of the ATIM window. Beacon generator handover occurs during the CBR Voice transmission schedule by specifying the new
Figure 2.1 indicates that the fixed lengths of the ATIM window and the beacon interval of IEEE 802.11 ad hoc power saving mode (PSM) often consume excess energy.
The inefficiency of the fixed ATIM window size under different traffic loads requires dynamic adjustment of the ATIM window size.
Finally, in the power saving mode (PSM) of IEEE 802.11, nodes that successfully announce ATIM frames stay active during the entire beacon interval. After Data transmission, nodes need not remain active state. After a node finishes Data transmission, the proposed power-saving MAC protocol switches the node from an active state to a sleep state to minimize power consumption. The novel feature of the proposed MAC protocol is the decreased idle time in active state and the switchover times (active-to-sleep state or sleep-to-active state).
The following sub-sections describe these features in more detail.
3.2 Basic Operation
The proposed power-saving MAC protocol matches time-frame duration to the periodic rate of voice packets. Figure 3.1 presents the frame format. Each frame consists of two sub-frames: a ATIM sub-frame and a data transmission sub-frame. The ATIM sub-frame consists of a beacon generation window and a ATIM window. The data transmission sub-frame consists of a CBR Voice window and Data window.
Figure 3.1: Representation of Proposed power-saving MAC frame format
3.2.1 Definition for new management frames in ATIM window
Because the ATIM and ATIM_ACK frame defined in the standard has MAC header, Frame Check Sequence (FCS) and an empty frame body, the standard was slightly modified to include that information in the ATIM and ATIM_ACK frame. A frame body of several octets was added to deliver that information can be easily done and will not affect other operations defined in the standard. Besides, the beacon frame carries the ROC (Reserved OCcupation Table) to indicate Reserved CBR Voice information. The following definitions and frame formats are used in the proposed protocol.
ROC (Reserved OCcupation Table)
– Indicates Reserved CBR Voice information including Source address, Destination address, and Transmission Order
– Beacon Frame carries ROC used by residential nodes
Data
BGS (Beacon Generation Sequence)
– Nodes with reserved CBR Voice must send Beacon with ROC at TBTT in sequence
– The beacon transmission sequence with ROC
• Every CBR Voice responder attached to Every CBR Voice sender to form the BGS.
Figure 3.2: Beacon Frame with ROC
Figure 3.3: ATIM and ATIM_ACK Frame for Data Time
001~110 for CBR Voice 0/1
3.2.2 Beacon generation procedure
In the initial startup stage, nodes at TBTT that have no BGS within the Ad Hoc contend to transmit the beacon frame (Fig. 3.4). Each node calculates a random delay uniformly distributed within the range from 1/4 to 2*CWmin*SlotTime (CWmin is the minimum contention window used in the backoff mechanism), then waits for the duration of the delay using the backoff mechanism. When the delay counter expires, the beacon is transmitted with a timestamp, which is a copy of the sender’s local TSF timer of the sender, with some adjustment. Nodes cancel the beacon transmission if a beacon arrives before their random delay counter expires. When a beacon is successfully received, stations set the TSF timer according to the timestamp of the beacon if the value of the timestamp is later than the TSF timer of the station itself in PSM of 802.11.
Therefore, the TSF timer can only be adjusted forward. In the proposed protocol, Timers can be set both forward and backward to ensure that all nodes wake up simultaneously.
Figure 3.4: Beacon generation Window
As soon as nodes have BGS, those that have reserved bidirectional CBR Voice must TBTT
1/4*CWmin*SlotTime 2*CWmin*SlotTime
B B B
Beacon without ROC Beacon with ROC
send a Beacon with ROC at TBTT in turn. If the Beacon generator node fails, the remaining network nodes should be able to compensate for this situation and should be able to continue normal operation as quickly as possible.The BGS is a more natural and complete way of backing up the beacon generator and it provides a perfect beacon transmission order list of ROC after ATIM exchange. Every CBR Voice responder is attached to CBR Voice sender to form the BGS. Nodes that have reserved CBR Voice can listen for the beacon and become the beacon generator whenever the previous beacon generator fails. The first sender in the sequence is the first generator; the second sender is the second generator; and the senders append the responders. The backup nodes listen to the beacon, which is a part of normal network operation. If the second generator does not hear the beacon for a short interframe space (SIFS) time, then the first generator is assumed dead and the second generator modifies the ROC and transmits the beacon with ROC. If the beacon is not transmitted for two SIFS times, then the third generator understands that both generators are dead. It then modifies ROC and transmits the beacon. If after N SIFS times no beacon is transmitted, then the rest of the nodes understand that No CBR Voice is reserved, and each node calculates a random delay uniformly distributed within the range between 1/4 and 2*CWmin*SlotTime and the transmitted beacon. The Maintenance scheme of Reserved OCcupation Table can be summarized as follows:
• Initially, nodes send ATIM message piggyback Transmission Order (s1, s2) if No
Beacon Generation Sequence (Fig. 3.5).
• If reservation is successful, pair nodes continues to reserve the Transmission in future cycles (Fig. 3.6).
• Finally, nodes send ATIM message and release the piggyback ROC after the conversation is completed or if the partner dies (Fig. 3.7).
• If the pair Nodes suddenly shut down, the other nodes should know the pair nodes have died and modifies ROC.
Figure 3.5: Maintenance scheme of Reserved OCcupation Table (一)
21
Figure 3.6: Maintenance scheme of Reserved OCcupation Table (二)
22
Figure 3.7: Maintenance scheme of Reserved OCcupation Table(三)
3.2.3 ATIM Information Exchange procedure
After the beacon transmission window, nodes enter the ATIM window. For ATIM information exchange, the system piggybacks the number of pending packets of ATIM announcement for Data. The information about the sender, receiver, and length of the pending data are contained in ATIM frame denoted ATIM (Sender ID, Receiver ID, Data Length). The transmission Release/Reservation status of ATIM announcement for CBR Voice is also piggybacked. The information about the sender, receiver, and Release/Reservation status of the CBR Voice traffic is contained in the ATIM frame denoted as ATIM (Sender ID, Receiver ID, TX status). According to IEEE 802.11 regulations, all nodes are fully connected and all PS mode nodes can wake up at almost the same TBTT. At the TBTT, each node wakes up for an ATIM window interval. If a node with buffering Data to a PS mode node, it sends an ATIM frame to the PS mode node within the ATIM window period. Upon receiving the ATIM frame, the PS mode node responds to an ATIM ACK to the sender of the ATIM frame and completes the reservation of the data frame transmissions. If a node with buffering CBR Voice to a PS mode node, it sends an ATIM frame to the PS mode node within the ATIM window period. Upon receiving the ATIM frame, the PS mode node responds with an ATIM ACK to the sender of the ATIM frame and joins the Reservation OCcupation Table (ROC). These pair nodes are not required to send further ATIM frames during the
or if their pair nodes do not respond to with an ATIM ACK to the sender during CBR voice transmission. Due to the broadcast nature of the wireless medium, each node can overhear the ATIM exchange information and get all node transmission tables (Data Length and CBR Voice Transmission Order) as Table 3.1 shows.
Each node employs prioritized contention-based Enhanced Distributed Channel Access (EDCA) as defined in the 802.11e [10]. Each transmission queue has a different interframe space (AIFS) and a different contention window limit. Each node that intends to transmit data calculates its priority according to the collected data profiles. The proposed protocol uses Data and voice traffic only. Nodes adjust the contention window size and AIFS time. T he smaller i t s AIFS, the higher t he priority a node can have.
Similar protocols apply to the contention window size. Figure 3.8 illustrates the time diagram of EDCA [10]. The 802.11e suggests the use of different AIFS and different contention window limits according to different ACs. Table 3 . 2 shows the parameters for the maximum contention window (CWmax), the minimum contention window(CWmin)
TX RX Data Length TX Order A C 10B s1 C A 15B s2 D C 30B
B D 20B s3 D B 10B s4 E B 25B
C E 40B
Table 3.1:Transmission table during ATIM window
and AIFS for each AC.
Figure 3.8: The timing diagram of 802.11e EDCA[10]
AC CWmin CWmax A
I
AC_BK aCWmin aCWmax 7
AC_BE aCWmin aCWmax 3
AC_VI (aCWmin + 1)/2 aCWmin 2
AC_VO (aCWmin + 1)/2 (aCWmin + 1)/2 2 Table 3.2: The default EDCA parameters [10]
The value of AIFS is determined by the following equation (1)[10]:
AIFS = AIFSN × aSlotTime + SIFS (1)
where the value of AIFS Number (AIFSN) is an integer greater than zero and is dependent on each AC.
3.2.4 Traffic-Load oriented ATIM window adjustment
To conserve the power used by PS mode nodes and to improve network throughput, the relationship between the lengths of the ATIM window and the corresponding beacon
interval is discussed in [7] as mentioned above. The authors indicated that the ATIM window
size should take approximately 1/4 of the beacon interval. The proposed scheme can
dynamically adjust the ATIM window according to the channel idle time and the actual traffic load of the network. Each node senses how long the channel is continuously idle during the ATIM window. If nodes sense that the channel is idle longer than AIFS (AC_BK) + CWmin, the assumption is that no nodes have buffered frames to send.
Besides, if ATIM collision occurs and Remaining Beacon Interval ≧ TMAXMPDU,
nodes could stay awake until no further ATIM collision occurs or remaining Beacon Interval ≦ TMAXMPDU. Therefore, all nodes can close the ATIM window and enter the
data transmission window according to transmission table. The following Fig. 3.9 is the traffic Load-oriented ATIM window adjustment algorithm.
Figure 3.9: Traffic-Load oriented ATIM window adjustment In ATIM window
Uses ATIM and ATIM_ACK with EDCA scheme to reduce collision and generate a contention-free transmission table
Dynamically Adjust ATIM Window Size {
While(1) {
R=0 R( Retransmit counter)
IF (Remaining BI <= TSIFS +TFRAME Min) Then AW=(Tcurrent-TBTT) ,Break;
ELSE Do {
CW=random [0, Min (CWmin*2R﹐CWmax) ]．SlotTime R=R+1
}
While( ATIM collision occur and Remaining BI ≧TMAXMPDU) IF (Tchannel idle ≧AIFS(AC_BK) + CW)
then AW=(Tcurrent-TBTT)+ AIFS(AC_BK) +CW , Break;
}
Nodes immediately transmit CBR Voice and Data on schedule }
3.2.5 Transmission Window operation
At the end of ATIM window, each PS mode node that successfully have finished ATIM exchange during an ATIM window wakes up to transmit its Voice/data packets and then enters a doze state according to its individual transmission order. Each node transmits CBR Voice according to the ROC and transmits data according to the Data transmission scheme.
To minimize the switchover penalty of data transmission, the heuristic algorithm developed by Prof .Tsern-Huei Lee et al. is used. The sub-optimum heuristic algorithm that minimizes the number of switchovers from the Doze state to the Awake state is determined based on graph theory. Details of the heuristic algorithm are shown below.
Heuristic Algorithm
Step 1. Set Schedule_List = ∅.
Step 2. Select a node n of graph G with minimum degree.
Step 3. Assume that Neighbor(n) = {n1, n2, …, n } and deg(j n )i ≤ deg(ni−1) for all i, 2≤ i ≤ j. Append the list of edges (n, n1), (n, n2), …, (n, n ) to Schedule_List j and then remove these edges from graph G.
Step 4. Remove a node if all edges incident on it are removed.
Step 5. Let G denote the resulting graph. Stop the algorithm if graph G becomes empty.
Step 6. While node n is removed, set n = j nj−1 till n = n1.
Step 7. If node n is removed, go to Step 2. Otherwise, go to Step 3.
The scenario in Fig 3.10 below is an example of bidirectional CBR Voice and data transmission operation.
Beacon
Figure 3.10: Example of our proposed protocol omitted the beacon generation procedure procedure
To simply the presentation, the beacon generation procedure in the previous section is omitted. Five PS mode nodes involved in the ATIM frame transmission are the following:
A,B,C,D and E. During the ATIM window, ATIM frames are announced successfully as follows, i.e., ATIM(A, C, 1),ATIM(B, D, 1), ATIM(A, C, 10), ACK(C,A, 15) ,ATIM(D, C, 30),ATIM(B, D, 20), ATIM(E, B, 25),ATIM(C, E, 40). Therefore, at the end of the ATIM window, all network nodes should maintain the transmission table in Tables 3.3
and 3.4 if no transmission errors occur.
Table 3.3: Example of bidirectional CBR Voice Table
After ATIM
Transmission Table
Table 3.4: Example of Data Transmission Table
After the ATIM window, firstly, nodes with bidirectional CBR Voice transmit voice packet according to ROC. The A and C wait a SIFS time, and A transmits a Voice packet according to ROC and C response ACK frame to A. At the same time, B and D wake up according to ROC. After a SIFS time, C transmits a Voice packet to A according to ROC.
The A then sends a response to ACK frame to C. After completing the exchange of bidirectional CBR Voice traffic, A and C go into a doze state. Then B continues to transmit bidirectional CBR Voice to D according to ROC. The D follows the same rule as A and C. After completing the bidirectional CBR Voice transmission, the nodes transmit data based on Data transmission order. At the same time, A and C wake up at the CBR Voice transmission time of D to B and A transmit data frames to C follows the Data transmission order. Other nodes then follow the same rule as A and C. Finally, B and D wake up according to Data transmission order at the transmission time of E to B and
transmit their data frames. After completing the Data transmission, B and D enter a doze state until the end of the beacon interval since there is no entry in their transmission table.
This chapter proposes a new protocol for improving throughput and power saving by dynamically adjusting ATIM window length. Nodes are also allowed to stay awake for only a fraction of the beacon interval following the ATIM window. Consequently, more stations can go into sleep mode in the middle of the beacon interval and stay in the sleep mode for a longer duration. Chapter 4 presents the results obtained by simulating the PSM scheme in IEEE 802.11, the proposed protocol, and power-saving mechanism proposed by the M. T Liu et al. [8] and the energy efficiency MAC protocol proposed by Wu et al. [9] in order to evaluate the performance of the proposed protocol.
Chapter 4
Simulation and Discussion
This chapter describes computer simulations to evaluate the energy conservation performance of the proposed protocol in the last section for an ad hoc wireless LAN consisting of ten to twenty hosts. The assumptions are that all hosts are fully-connected, and no transmission errors have occurred, implying that no packet is lost due to poor quality of the channel, and all protocol can avoid overhearing in data transmission window.
4.1 Simulation Model
For our simulation, C++ i s u s e d to implement a simulator The proposed MAC protocol is compared with the original IEEE 802.11 power-saving operation and the M.
T Liu et al. [8] proposed power-saving mechanism, and the Wu et al. [9] proposed energy efficiency MAC protocol. Two traffic types are modeled.
– Only Pure Data
– Pure Data and CBR Voice (6pairs in 10 nodes)
Each simulation was performed 100 times. A 2Mbps channel rate is assumed. To investigate non-real time and real-time traffic over an ad hoc network, two traffic types
are considered.
z Pure Data traffic: The arrival rate of data frames from each station is smaller than 10 kbits within 0.1sec. The frame size is 256 bits to 1024 bits long.
z Voice traffic: Voice traffic is usually considered constant bit rate (CBR) traffic.
After referring to other voice traffic models [11] , the data rate of voice traffic is set to 64Kbps. The value of maximum tolerance is 20 ms. Voice traffic is assumed to have highest priority, and pure data traffic has the lowest priority. Voice belongs to real-time traffic and has the maximum tolerance delay time. According to the traffic models defined above, Table 4.1 summarizes the traffic parameters of the protocol.
Table 4.1: Simulation Parameters
AIFSN=7,CWmin=15 Beacon Period (Voice Delay Bound )
ATIM window size
4.2 Performance Measurements
From many papers analysis on energy saving issues, some conditions are required be covered for each performance measurement:
1. Energy goodput: We used this energy goodput define in [12] to evaluate power efficiency.
Eergy goodput =
yConsumed TotalEnerg
ransmitted TotalBitsT
The unit of energy goodput is bits/Joule, this metric measures the amount of data delivered per joule of energy. The higher the energy goodput is, the lower the energy it consumes.
2. Throughput: The throughput is defined the channel rates that can be used to transmit pure frame by all traffics types of stations. The contention cost is excluded from the throughput.
Throughput =
TotalTime
lPackets fSuccessfu
theAmounto
* th Frame_Leng
3. Average Frame Delay: A period of the time from a frame arrives the system to it transmits completely.
4.3 Simulation Results
Figures 4.1 and 4.2 show the Energy goodput (bits/joule) for pure data and voice &
pure data traffic, respectively. Figure 4.1 shows the Energy goodput for four power-saving protocols with data traffic. When the number of nodes increases, the proposed protocol always achieves better energy goodput than any other protocol because it minimizes switchover time in data transmission window. The protocol consumes less power than other protocols do.
Figure 4.2 illustrates the Energy goodput for power saving protocols with voice &
pure data traffic. When the number of voice pairs increases, a node spends less time in doze state because of voice transmission. Because the proposed protocol provides a Voice reserved scheme to reduce contention among the nodes in ATIM window.
Figure 4.1: Number of Nodes v.s Energy goodput with pure data traffic
Figure. 4.2: Number of Voice pair v.s Energy goodput with pure data &Voice traffic
E n e r g y g o o d p u t
Figures 4.3 and 4.4 illustrate the average throughput (bits/sec/node) for pure data and voice & pure data traffic, respectively. Figure 4.3 shows the average throughput for four power saving protocols with data traffic. In the simulation results, PSM with ATIM
Figures 4.3 and 4.4 illustrate the average throughput (bits/sec/node) for pure data and voice & pure data traffic, respectively. Figure 4.3 shows the average throughput for four power saving protocols with data traffic. In the simulation results, PSM with ATIM