D3: Data-centric Data Dissemination in
Wireless Sensor Networks
Maarten Ditzel
Netherlands Organisation for Applied Scientific Research
P.O. Box 96864, 2509 JG Den Haag, The Netherlands
Email: maarten.ditzel@tno.nl
Koen Langendoen
Delft University of Technology
Mekelweg 4, 2628 CD Delft, The Netherlands
Email: k.g.langendoen@ewi.tudelft.nl
Abstract— This paper presents a novel method to disseminate
sensor data in a wireless sensor network, called D3 (Data-centric
Data Dissemination). The method combines the advantages of
data-centric routing like SPIN and directed diffusion and energy-
efficient MAC protocols such as S-MAC and T-MAC. The
protocol’s strengths are its energy-efficiency and its simplicity.
Messages are transmitted using broadcasting only, reaching as
many nodes as possible with the least energy. Furthermore, D3
easily accommodates energy-dependent traffic balancing and data
aggregation, crucial to prolong the lifetime of a sensor network.
I. INTRODUCTION
Wireless sensor networks [1], [2] have experienced increas-
ing attention in academic, industrial and military environments
over the past few years. These networks promise an easy-to-
deploy, easy-to-use and moreover, low-cost means to remotely
monitor environments. Furthermore, sensing accuracy can be
improved significantly by processing and combining collected
data within the network itself. Finally, the network can be
made robust to the failure of individual nodes, which ensures
that the lifetime and proper operation of the network is not
limited to the lifetime of one node in particular.
Applications, either envisioned or already realized, are
generally related to the remote monitoring of a, possibly
inaccessible or hostile, environment. Examples are an aqueous
surveillance system for a drinking water reservoir [3], a
wildlife habitat observation system [4], and a network to
monitor the behavior of glaciers [5]. Additionally, numerous
military applications are envisioned such as a battlefield data
collection network as described in [6].
In all the aforementioned applications, the network consists
of tens to millions of tiny devices. Each device carries one
or more sensors and has limited signal processing and com-
munication capabilities. Usually, the devices are powered by
batteries and can thus only operate for a limited time period.
Key to implementing a network with such devices is that
energy, computing power and communication bandwidth are
scarce. Therefore, light-weight, scalable, energy-conserving
communication protocols are essential to the successful oper-
ation of the network. Fast deployment of such a network and
robustness against device failures require an ad-hoc network
that is self-organizing. In general, radio communication (both
transmitting and receiving) is generally the operation that
consumes the most energy in a device.
Conventional ad-hoc address-oriented communication pro-
tocols, such as IEEE 802.11 [7], generally consume too much
energy or poorly support multi-hop networks. In [8] and [9] the
authors propose a new data-centric approach for the dissemi-
nation of data in sensor networks. At the same time, numerous
energy-efficient MAC protocols have been developed such as
S-MAC [10] and T-MAC [11]. However, these protocols still
operate in an address-oriented fashion. In accordance with
the shift in paradigm to data-centric communication protocols,
this paper proposes a novel method for communicating sensor
data through a wireless sensor network. It is a simple and
straightforward data-oriented method combining the ideas of
data-centric routing with an energy-efficient MAC protocol.
II. DATA-CENTRIC ROUTING
To operate a wireless sensor network successfully, two
closely related networking issues have to be addressed. The
first is the media access protocol, i.e., how the nodes access
and share the radio channel. The second is routing, i.e., how
data is passed through the network. Dependent on the structure
of the network different solutions may be found. In this paper
we consider a homogeneous, randomly distributed wireless
sensor network. All nodes are identical and offer the same
processing and communication capabilities. Several routing
schemes and MAC protocols can be found in literature (for
an overview see [12]).
Due to the energy restrictions on both processing and com-
munication, sensor data should be transmitted conservatively.
Based on this restriction, the authors of [8] suggest the use of
data-centric routing and introduced the family of the SPIN
protocols. Here, sensor nodes broadcast an advertisement
describing the available data and wait for a request of an
interested neighbor, before sending the actual data. In [9]
the authors propose the directed diffusion data dissemination
paradigm. A node, called the sink, transmits an interest,
describing the data it is interested in, to all nodes using
flooding. Subsequently, if new data becomes available, the data
is propagated using gradients that indicate where the interest
originated from.
Although both directed diffusion and the SPIN protocols
are data-centric routing protocols, they assume the use of an
address-centric MAC protocol, as they rely on message passing
to individual neighbors. As an alternative, this paper suggests
to use data-centric message passing instead.
III. D3: DATA-CENTRIC DATA DISSEMINATION
The D3 protocol is based on two basic assumptions. First,
transmitting information about data is cheaper in terms of
energy than transmitting the actual data itself.
1
Second, if
1
The same assumption is made in [8].
sink
0
1
1
1
2
2
3
INT
A
B
C
D
E
G
F
(a) Node A propagates an interest.
sink
0
1
1
1
2
2
3
ADV
A
B
C
D
E
G
F
(b) Node E advertises data.
sink
0
1
1
1
2
2
3
DATA
A
B
C
D
E
G
F
(c) Node E transmits the actual data.
sink
0
1
1
1
2
2
3
ADV
(vACK)
A
B
C
D
E
G
F
(d) Node D forwards the advertisement.
sink
0
1
1
1
2
2
3
DATA
A
B
C
D
E
G
F
(e) Node D forwards the data.
sink
0
1
1
1
2
2
3
(vACK)
(vACK)
ADV
A
B
C
D
E
G
F
(f) Node A acknowledges the data.
Fig. 1. The D3 protocol. After flooding an interest (INT), data is collected and forwarded using advertisements (ADV) to announce the transmission of the
actual data (DATA). Nodes uninterested simply ignore broadcasted data messages.
a node has new data available, at least one of its neigh-
bors is always interested.
2
Hence, the data must always be
transmitted. Therefore, a node does not require a confirmation
that a neighbor is interested and it can use broadcasting to
communicate the data. The neighbors can decide individually
whether to listen to the broadcast or not. Of course, these
assumptions have to be verified.
Following the basic principles of the T-MAC protocol [11],
the nodes adhere to an active-sleep regime. Periodically the
nodes wake up from sleeping and enter the active state. In the
active state nodes contend for the channel, applying collision
avoidance by means of an initial random contention delay
and carrier detection. During the sleep interval, messages are
queued. The active-sleep regime circumvents the problem of
idle-listening: the nodes are no longer required to keep their
radio in receive mode all the time, since they now know when
to expect a message from a neighbor.
The D3 protocol uses three types of messages to disseminate
sensor data:
• INT – interest. Interests are propagated from a node to all
other nodes in the network. It contains a unique identifier
and a task descriptor that describes what kind of data the
originating node (the sink) is interested in. Each sensor
node stores the interest and the hop distance to the sink,
also called the depth of the node with respect to the
particular interest.
• ADV – data advertisement. When a node has data that
suits a particular interest, it signals all its neighbors that
it has relevant data ready using the interest identifier and
its own depth. The data can either originate from its own
sensors, or it has been previously received by other nodes
and should be forwarded.
• DATA – data message. Data messages contain the actual
sensor data.
The basic operation of the D3 protocol is depicted in Fig. 1.
Node A starts with sending a data interest to all nodes using
flooding (Fig. 1(a)). When node E has new data available
(Fig. 1(b)) it advertises the data to its neighbors C, D, F, and
2
Sensor data is only collected after receiving an interest (similar to directed-
diffusion). Therefore, at least the neighbor the interest originated from must
be interested.
G, and claims a time slot for transmitting the data message.
Subsequently, it sends the actual data (Fig. 1(c)) at the reserved
time to all interested neighbors: C and D. Nodes F and G are
not interested in the data and simply ignore the transmission
by switching off their radios. This sequence of advertisement
and data transmission (Fig. 1(d) and Fig. 1(e)) is repeated to
forward the data until it reaches the sink where the interest
originated from. Forwarded ADV messages are also used as
virtual acknowledgements (vACK) (see section III-B). Finally,
the sink acknowledges the reception of the data (Fig. 1(f))
using a further unused advertisement as acknowledgement.
A. Detailed Operation
The sensor network starts at rest. No data is collected by
the nodes and no messages are communicated. To activate
particular nodes and sensors an interest is injected into the
network via a node, henceforth known as the sink. The
interest is propagated to all the nodes in the network using
flooding. Each node registers the interest’s identifier and its
corresponding depth. Once a node has received and registered
an interest, it inspects the task descriptor and activates the
corresponding sensors if necessary.
As soon as nodes have activated sensors they can start
collecting data. If a nodes collects data relevant to a stored
interest, it signals all its neighbors that it has data related
to a certain interest available. It does so by broadcasting
an advertisement containing the interest identifier, the node’s
depth with respect to the sink the interest originated from,
the start time of the data message and the length of the data
package. Furthermore, it attaches an identifier to the ADV
message. The node also stores the identifiers in a history, to
keep track of the recently sent ADV messages.
Upon receiving an advertisement, the node’s neighbors
check whether they are closer to the sink (lower depth) and,
if so, prepare to receive the data message by switching on
their radios at the specified start time. If not, they simply
ignore the upcoming transmission.
3
The same advertisement-
transmission sequence is used to forward data messages. As
3
In this paper a node’s depth is used as a decision criterion. However, other
cost metrics (e.g., the physical distance to the sink) can also be used.
data is transmitted only to nodes with lower depths, the data
will always propagate towards the sink node of the interest.
B. Virtual Acknowledgements
Apart from announcing that sensor data is available for
transmission, the ADV message serves another purpose,
namely as a virtual acknowledgement. In Fig. 1(d) node D
broadcasts an advertisement in order to forward data originally
coming from node E. When node E receives the broadcasted
ADV message, it will ignore the upcoming transmission as
usual, because the depth included in the advertisement is lower
than its own depth. However, it also checks the advertisement’s
identifier to see where the data originated from. If the node can
find this identifier in its own history of sent ADV messages, it
implicitly knows the data package has been properly received.
Hence, the ADV message also serves as an acknowledgement
of the proper reception of a data message. If the node does not
receive a virtual acknowledgment within a certain time after
transmitting a data message, it can assume the data is lost and
take appropriate actions.
Finally, the ADV message even has a third role. In case a
node receives an ADV message from a node with the same
depth (Fig. 1(d), node C), it will postpone the forwarding of
the corresponding data messages for a certain time, in order
to avoid duplication of data messages. However, if the node
does not duly receive a virtual acknowledgement (Fig. 1(f)),
it will reschedule the forwarding. In this way, energy waste
due to unnecessary transmissions is kept to a minimum, while
maintaining the flexibility to reach a sink along different paths.
IV. RESULTS
To verify the ideas and concepts introduced in this pa-
per, simulations have been carried out using the OMNet++
simulation environment [13] (see Fig. 2). Moreover, the D3
protocol is implemented on a small scale experimental sensor
network (see Fig. 3). Preliminary results show the proper
operation of the protocol and, especially, its scalability. Apart
from verifying the proper operation of the protocol, the first
experiments focused on possible redundant paths from the
source node to the sink, which may reduce the efficiency of
the protocol.
Redundant paths may arise when nodes fail to receive virtual
acknowledgements from neighboring nodes. Then a message
will be multiplicated and sent along different paths to the
sink. A worst case scenario is depicted in Fig. 4, where the
topology of the network is chosen such that every message is
duplicated. For clarity, only the transmission of ADV messages
is depicted.
Ideally, only one message should be received by the sink
for every message sent by the source, i.e., a redundancy of 1.
For multiple paths from the source to the sink, the redundancy
R is calculated using
R =
1
d(n
1
)
N
X
i=1
s(n
i
), (1)
where n
i
denotes the i
th
node, with n
1
the source node, s(n
i
)
the total number of data messages sent by node i, and d(n
1
)
the depth of the source node (e.g., in the example of Fig. 4
d(n
1
) = 4).
In the experiments the redundancy has been calculated for a
large number of randomly generated networks with increasing
Fig. 2. A screenshot of a running simulation using the OMNet++ simulation
environment.
connectivity. The connectivity of a network is the average
number of neighbors per node. Each network consisted of over
250 nodes. In Fig. 5, this redundancy is plotted against the
connectivity of the network. For each point in the plot, 100
simulations were run. In each simulation 100 messages were
sent from the source node to the sink. The figure shows only
a slight increase in the redundancy for networks with higher
connectivity.
Nevertheless, more extensive simulations are needed to
get sensible performance metrics in order to compare the
performance of the D3 protocol against other protocols such
as directed diffusion and the SPIN family of protocols in
combination with, for instance, T-MAC.
2.5 cm
Fig. 3. Picture of the TNOde platform.
V. CONCLUSION
In this paper, the shift in paradigm towards data-centric
routing is followed and incorporated into energy-efficient
MAC protocols. This leads to a novel data dissemination
protocol, called D3. The D3 protocol uses advertisement
messages to announce that new data is available. Interested
nodes listen to the subsequent transmission of the actual data,
while uninterested nodes simply ignore the data message. All
communication follows an active-sleep regime, alleviating the
problem of idle-listening.
ADV
sink
source
0
12
34
(a) The source transmits
a single message.
ADV
ADV
sink
source
0
12
34
(b) Duplicate messages
are sent.
ADV
0
12
34
ADV
ADV
ADV
sink
source
(c) Duplicated messages
are forwarded.
0
12
34
ADV
ADV
sink
source
(d) Duplicated messages
are combined.
ADV
sink
source
0
12
34
(e) The sink receives du-
plicated messages.
Fig. 4. As nodes along the vertical axes fail to receive each others advertisements, messages are duplicated as they propagate towards the sink node.
0 5 10 15 20 25
0
1
2
3
connectivity
redundancy
Fig. 5. The average number of redundant paths (connectivity) plotted against
a network’s connectivity.
The protocol combines the advantages of data-centric rout-
ing like SPIN and directed diffusion and energy-efficient
MAC protocols such as S-MAC and T-MAC. Its strength
is its energy-efficiency and its simplicity. Nodes only need
to process local data and control messages are kept to an
absolute minimum, both in number and in length. Furthermore,
D3 can be easily extended to accommodate energy-dependent
traffic balancing and data aggregation, vital for prolonging the
lifetime of a battery-powered network.
ACKNOWLEDGEMENT
The authors would like to thank Wouter Vlothuizen and Tom
Parker for their valuable comments and suggestions, which
greatly contributed in improving the protocol.
REFERENCES
[1] D. Estrin, R. Govindan, J. Heidemann, and S. Kumar, “Next century
challenges: scalable coordination in sensor networks,” in Proceedings
of the 5th annual ACM/IEEE international conference on Mobile com-
puting and networking. ACM Press, 1999, pp. 263–270.
[2] G. J. Pottie and W. J. Kaiser, “Wireless integrated network sensors,”
Commun. ACM, vol. 43, no. 5, pp. 51–58, 2000.
[3] X. Yang, K. G. Ong, W. R. Dreschel, K. Zeng, C. S. Mungle, and
C. A. Grimes, “Design of a wireless sensor network for long-term, in-
situ monitoring of an aqueous environment,” Sensors, vol. 2, no. 11, pp.
436–472, November 2002.
[4] A. Mainwaring, D. Culler, J. Polastre, R. Szewczyk, and J. Anderson,
“Wireless sensor networks for habitat monitoring,” in Proceedings of
the 1st ACM international workshop on Wireless sensor networks and
applications. ACM Press, 2002, pp. 88–97.
[5] “Glacsweb,” http://www.glacsweb.org.
[6] F. Ye, H. Luo, J. Cheng, S. Lu, and L. Zhang, “A two-tier data dissem-
ination model for large-scale wireless sensor networks,” in Proceedings
of the 8th annual international conference on Mobile computing and
networking. ACM Press, 2002, pp. 148–159.
[7] LAN MAN Standards Committee of the IEEE Computer Society, IEEE
Std 802.11-1999, Wireless LAN Medium Access Control (MAC) and
Physical Layer (PHY) specifications, IEEE, 1999.
[8] J. Kulik, W. Heinzelman, and H. Balakrishnan, “Negotiation-based
protocols for disseminating information in wireless sensor networks,”
Wirel. Netw., vol. 8, no. 2/3, pp. 169–185, 2002.
[9] C. Intanagonwiwat, R. Govindan, D. Estrin, J. Heidemann, and F. Silva,
“Directed diffusion for wireless sensor networking,” IEEE/ACM Trans.
Netw., vol. 11, no. 1, pp. 2–16, 2003.
[10] W. Ye, J. Heidemann, and D. Estrin, “An energy efficient MAC protocol
for wireless sensor networks,” in In proceedings of the 21st Conference
of the IEEE Computer and Communications Societies (INFOCOM),
vol. 3, June 2002, pp. 1567–1576.
[11] T. van Dam and K. Langendoen, “An adaptive energy-efficient MAC
protocol for wireless sensor networks,” in Proceedings of the first
international conference on Embedded networked sensor systems. ACM
Press, 2003, pp. 171–180.
[12] I. F. Akyildiz, W. Su, Y. Sankarasubramaniam, and E. Cayirci, “A survey
on sensor networks,” IEEE Communications Magazine, vol. 40, no. 8,
pp. 102–114, August 2002.
[13] “Omnet++: Objective modular network testbed in C++,”
http://www.omnetpp.org.
