DW-MAC: A Low Latency, Energy Efficient Demand-Wakeup MAC Protocol For Wireless Sensor Networks
DW-MAC: A Low Latency, Energy Efficient Demand-Wakeup MAC Protocol For Wireless Sensor Networks
DW-MAC: A Low Latency, Energy Efficient Demand-Wakeup MAC Protocol For Wireless Sensor Networks
Yanjun Sun
Shu Du
Omer Gurewitz
yanjun@cs.rice.edu
dushu@cs.rice.edu
gurewitz@cse.bgu.ac.il
David B. Johnson
dbj@cs.rice.edu
ABSTRACT
1. INTRODUCTION
Duty cycling is a widely used mechanism in wireless sensor networks (WSNs) to reduce energy consumption due to idle listening,
but this mechanism also introduces additional latency in packet delivery. Several schemes have been proposed to mitigate this latency,
but they are mainly optimized for light traffic loads. A WSN, however, could often experience bursty and high traffic loads, such as
due to broadcast or convergecast traffic. In this paper, we present
a new MAC protocol, called Demand Wakeup MAC (DW-MAC),
that introduces a new low-overhead scheduling algorithm that allows nodes to wake up on demand during the Sleep period of an
operational cycle and ensures that data transmissions do not collide at their intended receivers. This demand wakeup adaptively
increases effective channel capacity during an operational cycle as
traffic load increases, allowing DW-MAC to achieve low delivery
latency under a wide range of traffic loads including both unicast
and broadcast traffic. We compare DW-MAC with S-MAC (with
and without adaptive listening) and with RMAC using ns-2 and
show that DW-MAC outperforms these protocols, with increasing
benefits as traffic load increases. For example, under high unicast
traffic load, DW-MAC reduces delivery latency by 70% compared
to S-MAC and RMAC, and uses only 50% of the energy consumed
with S-MAC with adaptive listening. Under broadcast traffic, DWMAC reduces latency by more than 50% on average while maintaining higher energy efficiency.
General Terms
Algorithms, Design, Performance
Keywords
Sensor networks, medium access control, duty cycling, unicast
traffic, broadcast traffic, latency, energy
Permission to make digital or hard copies of all or part of this work for
personal or classroom use is granted without fee provided that copies are
not made or distributed for profit or commercial advantage and that copies
bear this notice and the full citation on the first page. To copy otherwise, to
republish, to post on servers or to redistribute to lists, requires prior specific
permission and/or a fee.
MobiHoc08, May 2630, 2008, Hong Kong SAR, China.
Copyright 2008 ACM 978-1-60558-073-9/08/05 ...$5.00.
53
R C DATA A
R C DATA A
C
C
Sync
R C DATA A
time
R C DATA A
Data
Transmit
Sleep
Receive
Active radio
2.
RELATED WORK
54
P P
P P P DATA A DATA A
PP
C
Sync
DATA A
time
S
T3
DATA A
Data
T1
Sync
Sleep
T2
wake up to transmit data packet
T4
Data
Sleep
TData
TSleep
T3
TData
T1
=
=
T2
T4
TSlee p
sensor nodes during the Sync period with required precision. The
basic concept of DW-MAC is to wake up nodes on demand during
the Sleep period of a cycle in order to transmit or receive a packet.
This demand wakeup adaptively increases effective channel capacity during a cycle as traffic load increases, allowing DW-MAC to
achieve low delivery latency under a wide range of traffic loads
including both unicast and broadcast traffic.
DW-MAC is unique in the way it schedules nodes to wake up
during the Sleep period of a cycle. In DW-MAC, medium access control and scheduling are fully integrated. In a Data period, a node with pending data contends for channel access using
a CSMA/CA protocol as in IEEE 802.11. DW-MAC, however, replaces RTS/CTS with a special frame called a scheduling frame
(SCH). The interval of time during which the transmission of a
SCH occupies the wireless medium automatically and uniquely reserves the proportional interval of time in the following Sleep period for transmitting and receiving the pending data packet. Essentially, DW-MAC sets up a one-to-one mapping between a Data
period and the following Sleep period. An SCH carries no timing information, and the transmission of an SCH simply replaces
that of RTS/CTS for medium access control. In this way, DWMAC minimizes scheduling overhead. As in an RTS, an SCH
contains the destination address, so this SCH wakes up only the
intended receiver, minimizing energy consumption due to unnecessary wake-ups. Furthermore, this integration ensures that data
transmissions do not collide at their intended receivers as discussed
below.
Figure 3 shows an overview of scheduling in DW-MAC based
on this one-to-one mapping between a Data period and the following Sleep period. In this example, node A wants to transmit a data
packet to node B. Node A first contends for channel access and
transmits an SCH during the Data period. Suppose transmission
of the SCH starts T1 time units after the beginning of the Data period. Based on T1 and the duration of the SCH transmission, T3 ,
nodes A and B will both schedule their wakeup time to T2 from
the beginning of the following Sleep period, and will agree on a
maximum wakeup duration of T4 , based on the ratio between TData
and TSleep , as shown in the figure. If the packet to be transmitted is a unicast packet, node B will return a confirmation SCH
frame (not in the figure) SIFS delay after receiving the request
SCH from A; if the packet is a broadcast packet, node B takes no
further action. When nodes A and B both wake up at the agreed
time, node A transmits the actual data packet, which can be either
broadcast or unicast. In case of unicast packet, node B acknowledges the successful receipt of the packet with an ACK. Although
we show the scheduling for only one pair of nodes in this example, DW-MAC allows multiple contending nodes to exchange SCH
frames with their intended receivers during a Data period, so that
multiple data transmissions can happen in the following Sleep period.
3. DW-MAC DESIGN
3.1. Overview
DW-MAC is a synchronized duty cycle MAC protocol, where each
cycle is divided into three periods: Sync, Data, and Sleep (Figure 3). We denote the duration of each period by TSync , TData , and
TSleep , respectively. Similar to prior work, DW-MAC assumes that
a separate protocol (e.g., [6, 9]) is used to synchronize the clocks in
55
T1D
B
Sync
T1S
DATA
DATA
Data
Sleep
56
T1D
T2S
SS
Sync
S S
S
Sync
T1D
T1S
SS
T2D
S S
C
Sync
Data
DATA A
Data
T3S
DATA
DATA
Sleep
DATA A
T2S
SS S
DATA
T3D
DATA
T1S
S S S
Sleep
DATA
T1D
DATA A
Data
T2S
T2D
DATA A
T2D
SIFS
T1S
SS
DATA A
DATA A
Sleep
mized forwarding makes it possible for an SCH and thus the corresponding data packet to reach further nodes in a single cycle than
having all rebroadcasting nodes compete for the medium equally.
Although this reduced randomness could increase collision probability, the improved spatial reuse usually offsets this increase or
even lowers total collision probability, as we reveal in our evaluation in Section 4. Many criteria can be used for choosing an immediate forwarder, such as location, degree, or the number of children
nodes of a neighbor. In our DW-MAC simulation, this optimized
forwarding is used when a broadcast tree of a WSN is available
and a node knows its children nodes height (the number of edges
on the longest downward path to a leaf). For an SCH to be rebroadcast, if the SCH is received from the parent node, a node chooses
the child with greatest height as the immediate forwarder. If this
SCH is received from one child node, the parent node of the SCH
receiver is chosen as the immediate forwarder.
57
20 Kbps
31.2 mW
22.2 mW
22.2 mW
3 W
5 ms
10 ms
5
5%
S-MAC
RMAC
DW-MAC
4.
SIMULATION EVALUATION
58
150
100
95
100
90
SMAC
SMACAL
RMAC
DWMAC
85
80
50
75
0
100
150
200
450
500
70
100
200
300
400
Sensing Range (m)
500
SMAC
SMACAL
RMAC
DWMAC
Delivery Ratio
200
2.5
SMAC
SMACAL
RMAC
DWMAC
1.5
0.5
100
200
300
400
Sensing Range (m)
500
ated for each event with the 500-meter sensing range, DW-MAC
reduces average end-to-end delay by around 70% compared to SMAC and RMAC. DW-MAC outperforms S-MAC because DWMAC allows more transmissions in a cycle by using the Sleep period for actual data transmissions. RMAC experiences more delay
than DW-MAC as workload increases, because of increased packet
collisions caused by scheduling conflicts. It is the retransmission
effort to recover these collided packets that results in larger end-toend delay. When the sensing range is 500 meters, the maximum
end-to-end delay with RMAC is 374.95 seconds, which is off the
top of the graph. This extreme delay occurs when a packet generated for one event failed to reach the sink before the next event
happened. Under the light traffic with the 100-meter sensing range,
DW-MAC shows slightly larger delay than RMAC, due to the time
that a received data packet is forwarded to the next hop in multihop
forwarding. In RMAC, a data packet is forwarded immediately,
whereas in DW-MAC, forwarding starts at a later time determined
by the corresponding SCH frame. This extra delay experienced by
DW-MAC, however, is less than the duration of a Sleep period. SMAC with adaptive listening shows slightly larger delay compared
to DW-MAC. This low delay achieved by adaptive listening, however, comes at the cost of lower packet delivery ratio and increased
energy consumption as shown next.
The packet delivery ratios corresponding to Figure 8(a) are
shown in Figure 8(b). DW-MAC maintains close to 100% packet
delivery ratio and outperforms the other protocols across all sensing ranges. The delivery ratio with S-MAC with adaptive listening
drops quickly, since with larger the sensing ranges, more collisions
are caused by transmissions from hidden nodes, as we discussed in
Section 2; in addition, a node may transmit a packet when its intended receiver is in sleep state, further decreasing packet delivery
ratio. DW-MAC and RMAC outperform S-MAC mainly for two
reasons. First, they only transmit short scheduling frames during
a Data period, avoiding collisions between a control frame and a
long data frame. Second, a node does more retransmission attempts
for a data packet in DW-MAC and RMAC. Specifically, a scheduling frame sent by an intermediate node in multihop forwarding
serves both as RTS and as CTS; even if this frame fails to reach
the next-hop neighbor, the intermediate node does not increase its
59
Cumulative Fraction
Cumulative Fraction
0.6
0.4
SMAC
SMACAL
RMAC
DWMAC
0.2
0
0
50
100
150
200
EndtoEnd Delay (s)
250
0.8
Cumulative Fraction
0.8
0.8
SMAC
SMACAL
RMAC
DWMAC
0.6
0.4
0.2
0.4
SMAC
SMACAL
RMAC
DWMAC
0.2
0
88
300
0.6
90
92
94
96
Delivery Ratio
98
0
0
100
1
2
3
Average Energy Consumption of Sensors (mW)
80
SMAC All
SMAC CDS
DWMAC All
DWMAC CDS
DWMAC CDSMH
60
100
95
Delivery Ratio
70
50
40
30
90
85
80
SMAC All
SMAC CDS
DWMAC All
DWMAC CDS
DWMAC CDSMH
20
75
10
0
5
7
9
Number of Nodes along Edge of Grid
11
70
5
7
9
Number of Nodes along Edge of Grid
11
Figure 9: Performance for random correlated-event traffic in 50-node networks with sensing range of 250 m.
2
1.5
SMAC All
SMAC CDS
DWMAC All
DWMAC CDS
DWMAC CDSMH
0.5
5
7
9
Number of Nodes along Edge of Grid
11
a packet than does a non-CDS node with DW-MAC, due to defers caused by undecodable frames. When a node fails to decode
a received a packet, it defers for some time (such as EIFS in IEEE
802.11) to avoid interrupting ongoing transmission. Since this defer is much shorter than a Sleep period in our simulations, all neighboring nodes still compete for the medium fairly at the beginning
of the next cycle with S-MAC. With DW-MAC, however, it is possible that a node is ready to rebroadcast a packet before its defer
timer expires, as multiple SCHs can be transmitted during a Data
period. A CDS node that defers could be slower in rebroadcasting
a packet compared to a non-CDS node that does not defer, resulting
in lower latency for DW-MAC All than for DW-MAC CDS. However, DW-MAC still reduces end-to-end latency by around 40% for
CDS-based flooding compared to those with S-MAC.
Figure 10(b) shows the delivery ratios (the percent of broadcast
packets that are successfully received by all nodes in a network)
of flooding in the grid networks. Because of the increased redundancy in simple flooding, S-MAC and DW-MAC achieve higher
delivery ratio than CDS-based flooding. In simple flooding, DWMAC outperforms S-MAC, since the use of (short) SCH frames
instead of long data packets during contention helps to avoid collisions. However, when CDS-based flooding is used, DW-MAC
sometimes shows lower delivery ratios than does S-MAC, mainly
due to the special grid topology and selection of CDS as discussed
before. Looking at the results in random networks (Figure 11(b)),
on average, DW-MAC shows better delivery ratios than S-MAC
when CDS-based flooding is used. With improved spatial reuse
when optimized multihop forwarding is used, DW-MAC achieves
higher delivery ratios than does S-MAC in CDS-based flooding.
Average energy consumption in the grid networks, calculated as
we did in evaluations under unicast traffic, is shown in Figure 10(c).
60
0.8
0.6
0.4
SMAC All
SMAC CDS
DWMAC All
DWMAC CDS
DWMAC CDSMH
0.2
20
40
60
EndtoEnd Delay (s)
80
100
Cumulative Fraction
Cumulative Fraction
0.8
0
0
1
SMAC All
SMAC CDS
DWMAC All
DWMAC CDS
DWMAC CDSMH
0.8
Cumulative Fraction
0.6
0.4
0.6
0.4
0.2
0.2
0
50
SMAC All
SMAC CDS
DWMAC All
DWMAC CDS
DWMAC CDSMH
60
70
80
Delivery Ratio
90
100
0
0
0.5
1
1.5
Average Energy Consumption of Sensors (mW)
5.
ACKNOWLEDGEMENTS
We thank Jingpu Shi for valuable discussions during the design of
DW-MAC. This work was supported in part by the U.S. National
Science Foundation under grants CNS-0520280, CNS-0435425,
CNS-0338856, and CNS-0325971; and by a gift from Schlumberger. The views and conclusions contained here are those of the
authors and should not be interpreted as necessarily representing
the official policies or endorsements, either express or implied, of
NSF, Schlumberger, Rice University, Ben Gurion University, or the
U.S. Government or any of its agencies.
REFERENCES
[1] G. Anastasi, A. Falchi, A. Passarella, M. Conti, and
E. Gregori. Performance Measurements of Motes Sensor
Networks. In Proceedings of the 7th ACM International
Symposium on Modeling, Analysis and Simulation of
Wireless and Mobile Systems (MSWiM 2004), pages
174181, Oct. 2004.
[2] Qing Cao, Tarek Abdelzaher, Tian He, and John Stankovic.
Towards Optimal Sleep Scheduling in Sensor Networks for
Rare-Event Detection. In Proceedings of the Fourth
International Symposium on Information Processing in
Sensor Networks (IPSN 2005), pages 2027, Apr. 2005.
[3] Chipcon. Single Chip Very Low Power RF Transceiver
(CC1000 Datasheet), Apr. 2002.
[4] Tijs van Dam and Koen Langendoen. An Adaptive
Energy-Efficient MAC Protocol for Wireless Sensor
CONCLUSION
61
[5]
[6]
[7]
[8]
[9]
[10]
[11]
[12]
[13]
[14]
[15]
[16]
[17]
62