Study of Timing Properties on Data Collection and Query Processing
Techniques in Wireless Sensor Networks
Abderrahmen Belfkih, Bruno Sadeg, Claude Duvallet and Laurent Amanton
University of Le Havre, BP 1123, 76063 Le Havre cedex, France
Keywords:
Wireless Sensor Network, Data Collection, Query Processing, Sensors Database, Time Constraints.
Abstract:
The real-time aspect is an important factor for Wireless Sensor Network (WSN) applications, which demand a
timely data delivery, to reflect the current state of the environment. However, most existing works in the field
of sensor networks are only interested in data processing and energy consumption, to improve the lifetime of
WSN. Extensive researches have been done on data processing in WSN. Some works have been interested in
improving data collection methods, and others have focused on various query processing techniques. In recent
years, the query processing using abstract database technology like Cougar and TinyDB have shown efficient
results in many researches. This paper presents a study of timing properties through two data processing
techniques for WSN. The first is based on a warehousing approach, in which data are collected and stored on a
remote database. The second is based on query processing using an abstract database like TinyDB. This study
has allowed us to identify some factors which enhance the respect of temporal constraints.
1 INTRODUCTION
Wireless sensor networks can be considered as a type
of ad hoc wireless networks. They are composed
of small wireless nodes, which are manually or ran-
domly deployed in a region of interest, to sense differ-
ent physical characteristics of the environment, like
temperature, humidity, pressure, light, and so on. The
nodes transmit the sensed data to a sink node referred
to as Base Station (BS).This process, called data col-
lection, must be able to meet certain deadlines, i.e. to
update data before the expiration of their validity in-
tervals, in order for the WSN reflects the current state
of the environment.
WSN applications, such as industrial process
monitoring and control, can be considered as time-
critical. They require strict deadlines to send data to
the sink nodes. WSN are often deployed without giv-
ing much importance to the temporal constraints in-
cluding the data freshness constraint and the deadline
constraints.
Data freshness constraint imposes a limit on the
data validity period from the time it is read from a
sensor. The constraint of data deadline also requires
that data have to be collected within a deadline (Bhat-
tacharya, 2008). In real-time applications, data fresh-
ness is a key performance factor. It is important
to properly study the efficiency of data management
mechanisms in WSN.
Nowadays, researchers are interested in data pro-
cessing techniques to offer an efficient solution to in-
crease the WSN lifetime. They have proposed new
techniques for data aggregation, data collection and
query processing, to simplify the data extraction and
management in WSN. The query processing systems
such as Cougar (Fung et al., 2002) and TinyDB (Mad-
den et al., 2003) use the abstract database approach
which considers the WSN as a relational database.
They aim to improve the data extraction mechanisms
in WSN to save energy and to reduce the communica-
tion losses. (Garc´ıa-Hernando et al., 2008).
In this paper, we are interested in investigatingtwo
approaches for data processing in WSN (i) a query
processing approach using an abstract database and
(ii) a periodic data collection. We simulate temporal
constraint in order to identify which method givesbet-
ter data arrival time to the user and that ensures data
freshness in WSN. We consider the following perfor-
mance parameters: (i) the network convergence time,
which is the time that a sensor node takes to connect
to the base station, to build and to update its routing
tables. It can affect the arrival time of data, (ii) the
data collection time. The fastest is this parameter, the
more fresh are data, (iii) the database response time,
which can improve the data arrival time to the user.
The remainder of this paper is structured as fol-
77
Belfkih A., Sadeg B., Duvallet C. and Amanton L..
Study of Timing Properties on Data Collection and Query Processing Techniques in Wireless Sensor Networks.
DOI: 10.5220/0005273200770084
In Proceedings of the 4th International Conference on Sensor Networks (SENSORNETS-2015), pages 77-84
ISBN: 978-989-758-086-4
Copyright
c
2015 SCITEPRESS (Science and Technology Publications, Lda.)
lows. Section 2 reviews related work. In Section
3, we describe the two approaches cited previously.
In Section 4, we carry out simulations, thereafter we
present and comment the results obtained. Finally,
in Section 5, we conclude the paper by showing how
to effectively exploit the two techniques and we give
some perspectives to this work.
2 RELATED WORK
In this section, we give an overview of some research
works, which have studied the time criterion for WSN
data processing. Thereafter, we present the different
abstract databases for sensor networks.
2.1 Data Freshness in WSN
Suriyachai et al. have discuss the time-critical data
delivery in WSN (Suriyachai et al., 2010). They
have presented a novel TDMA-based Medium Ac-
cess Control (MAC) protocol named GinMAC, which
incorporates routing, topology control and reliability
control mechanisms. It aims to ensure timely data
delivery and the reliability control mechanism, with
minimum energy consumption in WSN. The authors
have shown that GinMAC is able to balance delay, re-
liability and energy consumption requirements for the
system with time-critical data delivery in WSN.
Sharaf et al. have presented a new approach based
on SQL-Like query syntax, named TiNA (Temporal
Coherecy-Aware in-Network Aggregation) (Sharaf
et al., 2003). TiNA exploits temporal coherency tol-
erance, to reduce the number of transmitted messages
by each node. The authors have proposed a new
clause named ”tct” (temporal coherencytolerance) for
each query, in order to express temporal coherency
tolerance. For example, the following query involves
getting the total light for rooms and grouping by floor,
with tct = 10%. The new reading is not reported if the
difference between it and old one is smaller than the
associated tct value.
SELECT {FLOOR, SUM(LIGHT)}
FROM SENSORS
GROUP BY {FLOOR}
EPOCH DURATION 30s
VALUES WITHIN 10%
Choi et al. consider that the aggregation time in WSN
can be affected by the deadline constraints, i.e, data
validity period or data delivery deadline (Choi et al.,
). They have proposed a new aggregation time control
(ATC) algorithm, which adjusts the aggregation time
of the node to ensure data delivery without exceeding
fixed deadlines.
Hiromori et al. have proposed a new protocol,
named DD (Drainage Divide), for periodic data col-
lection in WSN multi-sinks (Hiromori et al., 2012).
DD transforms the network to a set of trees, rooted at
the sinks. Then, it adjusts them, in order that the de-
lay from each node to the corresponding sink, meets
given deadlines. The authors have also applied the
CSMA/CA-based MAC delay analysis to estimate the
transmission delay at each node. DD protocol has
guaranteed data delivery within time, and it has sat-
isfied the given delay constraints in WSN.
There are many research efforts to study temporal
constraints such as delay constraints, deadline con-
straints and data delivery deadline on data processing
techniques like data aggregation, data collection and
query processing. However, there are few existing
researches works which focus on studying temporal
constraints in query processing systems, also called
abstract databases.
2.2 Abstract Database for WSN
Abstract databases for WSN have been cited for the
first time by Bonnet et al. (Bonnet et al., 2001). They
consider the WSN as a relational database that can be
queried using SQL-like queries. Abstract databases
present a declarative approach to facilitate the de-
scription and processing of queries in the WSNs. Data
required by the base station form a virtual table, in
which columns rerepresent the data requested by the
user.
Madden et al. have presented a new query pro-
cessing system, named TinyDB, using TinyOS (Levis
et al., 2005) as an operating system, for extracting
information from a network. It is composed of two
subsystems: (i) TinyDB application, running on each
node of the network, (ii) Java-based client interface,
allowing the user to describe the data using a declar-
ative SQL-LIKE queries, i.e., without knowing how
the data are processed in network. The users writes
the query that will be compiled and split into tasks,
then injected into the network. Once a node receives
the query, it processes it, and returns the results to the
user via the base station. For example, to calculate the
average temperature for all sensors when the light is
greater than 400 for a period of 10 seconds, we write
the following query:
SELECT AVG (temperature)
FROM sensors
WHERE light > 400
EPOCH DURATION 10s
TikiriDB (Laxaman et al., 2010) is a database ab-
straction layer for ContikiOS. It provides function-
SENSORNETS2015-4thInternationalConferenceonSensorNetworks
78
ality to collect data from WSNs without having any
information about programming knowledge of sensor
nodes. TikiriDB provides a TikiriSQL library, to re-
ceive queries from client, to parse them and to gen-
erate query packets. The results are sent to the gate-
way computer via the Serial Forwarder (SF) applica-
tion. The node connected to the gateway, receives the
queries, then injects them in WSN. It receives there-
after the results, which will be forwarded to the end
user by the SF application. In the following, here is
an example of TikiriDB query which permits to get
the light and the temperature from all sensors whose
temperature is above50. Each new reading will be de-
livered to the end user every 2 seconds for a duration
of 10 seconds.
SELECT light,temperature
FROM sensors
WHERE temperature > 50
SAMPLE PERIOD 2s FOR 10s
Amato et al. have proposed MaD-WiSe (MAnage-
ment of Data in WIreless SEnsor networks) (Amato
et al., 2010), which considers a WSN as a highly dis-
tributed and dynamic database. MaD-WiSe takes ac-
count of various aspects related to database system
design. Users can express queries using an SQL-like
language, named MW-SQL, to collect, to filter, and to
manage sensors data.
Corona (Khoury et al., 2010) is a distributed in-
network query processing system which provides a
declarative query language SQL-Like to formulate
queries. Corona introduces the notion of freshness
into WSN, allowing the user to obtain data from a sen-
sor network with data freshness guarantees. It uses
multiple aggregate queries to reduce processing de-
lays and costly communication in WSN. It is capable
of running multiple applications simultaneously.
An example of Corona query is given in the fol-
lowing:
SELECT temperature
FROM sensors
WHERE light > 100
EPOCH 60s
FRESHNESS 10s
In this example, the user queries all sensors to get the
temperature greater than 50 for epoch period of 60
seconds. It defines also the freshness constraint clause
of 10 seconds to fix the time allowed between two
acquisitions.
We note that temporal constraints have not been
considered in the most of abstract databases. They
have been studied only in the recent Corona work
(Khoury et al., 2010) . In the next section, we il-
lustrate our network model through two scenarios: a
periodic data collection and a query processing using
TinyDB and we present the parameters we selected to
discuss them in section 4.
3 NETWORK MODEL
In the two scenarios, we have used a network model
composed of four components: (i) sensor nodes
which report informations about the environment, (ii)
a base station which collects periodically data from
sensor nodes, (iii) a remote database used to store
the received data, (iv) a user which can connect to
the database and get information about WSN. In the
second scenario, we have added an abstract database
for WSN, named TinyDB, which allows the user to
send queries to the base station and to get queries re-
sponses.
3.1 First Scenario: Data Collection with
Remote Database
In this scenario, sensor nodes are randomly deployed
over an area and the base station is placed outside the
network. Nodes communicate amongst themselves
through wireless communications. Sensor nodes re-
trieve data values like temperature and humidity, and
send them to the base station. The data received are
transmitted to the remote database to be stored. Fi-
nally, they can be displayed through an interface to
the users.
Sensor network
User 3
Database server
Queries
Base Station
Data coll
ection
User 1
Data logging
User 2
response
Figure 1: Data collection with remote database.
3.2 Second Scenario: Query Processing
with WSN Abstract Database
This scenario is based on a query processing tech-
nique, in which we use an abstract database system,
TinyDB. User specifies the data it want to collect from
WSN, through SQL-like declarative queries. It sends
StudyofTimingPropertiesonDataCollectionandQueryProcessingTechniquesinWirelessSensorNetworks
79
Sensor network
User 3
Database server
Queries
Base Station
Data coll
ection
User 1
User 2
Queries
Queries
Abstract database
TinyDB
Data logging
response
response
response
Figure 2: Query processing with abstract database.
them to the base station via the the abstract database
interface. The base station will inject these queries
over the network. Each sensor receives a query, re-
sponds with the values requested for a duration fixed
in the query by the user. The base station sends the
responses to the user via the abstract database system
and they can be stored in a remote database.
We have considered some performance parame-
ters for the two approaches and we discussed them in
next section. These parameters are: (i) the network
convergence time, an important factor to ensure the
temporal constraints in WSN. It precedes the process
of data collection. It can affect on the arrival time of
data when it takes a long of time. We consider also
(ii) the data collection time, which presents an im-
portant factor to guarantee the freshness of data and
the data delivery deadline. When the data collection
is fast, data freshness will be better. Finlay, (iii) the
database response time which can be considered as an
important factor in the first scenario, because it repre-
sents the only data available to the user. We will show
in the next section (simulations) that sometimes, data
freshness cannot be met in this stage.
In the next section, we implement the two scenar-
ios described above and we discuss the performance
parameters for the two approaches. We identify and
we analyze also different impacts which can affect
these parameters.
4 SIMULATION & RESULTS
4.1 Simulation Environment
4.1.1 First Scenario
We have used COOJA (
¨
Osterlind et al., 2006), the
network simulator for ContikiOS, to create the first
scenario. We have created one sink node as base sta-
tion and a set of sender nodes of type Tmote Sky with
48 kb of flash memory. The nodes are randomly dis-
tributed in a area within 100m*100m and the sink is
placed outside of the deployment area. The number
of sensors vary from 10 to 100 with step of 10. Each
node sends temperature and humidity values to the
sink node every 60 seconds. Table 1 summarizes the
simulation parameters.
Table 1: Simulation Parameters.
Parameters Values
Operating system ContikiOS 2.6
Simulator Cooja
Mote type Tmote Sky
Simulation area (m) 100x100
Number of nodes 10-100
Number of sink Nodes 1
Source data rate (events/s) Random
Radio range (m) 80
Data packet size (bytes) 15
We use RPL (Routing Protocol for Low-Power
and Lossy Networks) (Winter et al., 2012) to provide
communication between sensor nodes and the base
station. RPL is a routing protocol designed for low
power and lossy networks and for large networks. It is
more efficient for data delivery time than other exist-
ing known protocols such as LOAD (LOADis derived
from AODV), DSR and DSDV (Vucinic et al., 2013).
It provides efficient routing paths guaranteeing data
delivery before deadline in WSN. We have used Con-
tikiMAC radio duty cycling protocol (Dunkels, 2011)
inspired from existing duty cycling protocols. It has
shown efficient time management because it uses a
wake-up mechanism to defines a precise timing be-
tween packet transmission. We have used the default
Contiki MAC layer CSMA (carrier sense multiple ac-
cess) which is the only one MAC driver in Contiki.
The table 2 below summarizes the different layers
protocol used in simulation.
Table 2: Simulation layer protocol parameters.
Transport layer UDP
Network layer IPV6/RPL
Radio duty cycling ContikiMAC
MAC layer CSMA
Physical layer IEEE 802.15.4
4.1.2 Second Scenario
We used TinyDB as abstract database in TinyOS.
TinyDB runs on Mica2 motes with 128 KB flash
memory because its code execution requires a flash
memory upper than 48 KB. We have chosen to use
SENSORNETS2015-4thInternationalConferenceonSensorNetworks
80
TinyDB because it is the most widely used system
by researchers and it has shown its effectiveness in
maintaining energy and for data managing. The net-
work is based on Mica2 sensors randomly scattered
in the field. We have varied the number of sensors
from 10 to 100 by step of 10. The sensor are queried
on demand by the user via the base station. Table 3
summarizes the simulation parameters.
Table 3: Simulation Parameters.
Parameters Values
Operating system TinyOS 1.x
Simulator Tossim
Mote type Mica2
Simulation area(m) 100x100
Number of sender nodes 10-100
Number of sink Nodes 1
Radio range (m) 80
Data packet size (bytes) 20
TinyDB uses a tree-Based Routing Protocol
(Gehrke and Madden, 2004) to provide the query dis-
semination and the result collection. Each node for-
wards the query to other nodes in the network, to form
the routing tree. This process ends when all nodes
will receive the query. TinyDB uses CSMA proto-
col, which is the already existing default MAC pro-
tocol used by TOSSIM. It uses also the Low power
listening (LPL) for low power communication and
the Drip/Drain routing. Drip is a transport-layer
component for disseminating messages throughout a
network and Drain is a collection routing layer for
TinyOS 1.x. Table 4 below summarizes the different
layers protocol used in simulation.
Table 4: Simulation layer protocol parameters.
Transport layer Drip/Drain
Network layer Tree-based routing protocols
Radio duty cycling Low-Power Listening
MAC layer CSMA
Physical layer IEEE 802.15.4
4.2 Simulation Description
4.2.1 First Scenario
Each sender node uses the UDP sender algorithm to
send its data to the base station. We have defined a
transmission period of 60 seconds. We have chosen
a random waiting time (SEND
TIME), before send-
ing packets to avoid that the senders sending pack-
ets at the same time. We have activated the digital
sht11 sensor, relative to sensor Humidity and tem-
perature. A sender node listens to its neighbors us-
ing tcpip
handler() function and continues to send
data currently detected to the base station using the
”send
packet” function.
Algorithm 1: UDP sender process.
1: define period 60
2: define send time random Value
3: sensor activate(sht11 sensor)
4: while 1 do
5: if ev = tcpip
event then
6: tcpip
handler()
7: end if
8: if etimer
expired(period) then
9: etimer reset(period)
10: ctimer
set(backoff time, send time,
11: send packet, null)
12: end if
13: end while
4.2.2 Second Scenario
When TinyDB receives the query from the user, it
parses and translates it to SQL query into an execu-
tion plan. Then, it injects it into the network via the
base station. Each node receives the query, processes
it using a query processor and returns the requested
values. When a result arrives, TinyDB calls the result
listener method and displays the query result for the
user using a Java-based client interface.
4.3 Simulation Results
4.3.1 Impact of Number of Nodes on Network
Convergence Time
The network convergence time is defined as the time
needed for the sensors to connect to the base station
and to build routing tables. This step comes just be-
fore the process of data collection. It can affect the
arrival time of data. The shorter is the convergence
time, the quicker is the availability of the data. In Fig-
ure 3, the network convergence time increases when
the number of nodes increases. The network conver-
gence time depends on the DODAG building process,
which begins at the root (base station). The root starts
the dissemination of information about the structure
using DIO (DODAG Information Object) message.
The nodes not connected to the tree (without commu-
nication with the root) receive the message DIO and
process it. Then they decide to join or not the struc-
ture. Once the node has joined the structure, it has
a route to the root of the DODAG structure. Build-
ing the final routing table depends on the number of
nodes and the path cost between nodes. We note also
StudyofTimingPropertiesonDataCollectionandQueryProcessingTechniquesinWirelessSensorNetworks
81
0
100
200
300
400
500
0 10 20 30 40 50 60 70 80 90 100
Time (second)
Number of nodes
Data collection
Figure 3: Network convergence time for data collection.
0
2
4
6
8
10
12
0 10 20 30 40 50 60 70 80 90 100
Time (second)
Number of nodes
Query Processing
Figure 4: Network convergence time with TinyDB.
in Figure 4 that the convergence time for TinyDB (ac-
quisition on demand) is less than that of data collec-
tion (periodically). TinyDB constitutes a routing tree
or Directed Acyclic Graph (DAG) with all sensors in
which the root is the sink (gateway). TinyDB uses a
selecting multi-hop tree based on efficiency of rout-
ing which depends on energy consumption. Sensors
make intelligent routing decisions, where every node
selects a parent based on link quality to the base sta-
tion. The nodes keep both a short list of neighbors
which have a recently heard transmit, and some rout-
ing information about the connectivity of those neigh-
bors to the rest of the network. They associate also a
link quality with each of their neighbors.
4.3.2 Impact of Number of Nodes on Data
Collection Time
Data collection time represents the time taken to get
the responses from all sensors connected to the base
station. It can affects the data validity time and then
not reflecting the current state of the environment.
The time required to send the data from sensor to the
base station depends on the sensors capacity, its po-
sition and the path quality depend on the number of
hops to the base station. Generally, the sender uses
multi-hop paths to send its data when the base station
is far located, which leads to transmission delay and
can cause a failure, i.e. to not respect the data validity
time.
0
100
200
300
400
500
600
0 10 20 30 40 50 60 70 80 90 100
Time (second)
Number of nodes
Data collection
Query processing
Figure 5: Completed Cycle Time.
Figure 5 shows the time spent by the nodes to
complete the sending of their data, using both data
collection technique and query processing technique.
We observe that the curve increases when the num-
ber of sensors increases for the two techniques. The
time required to send the data depends of the number
of hops and the availability of parent nodes to send
data received from children. Their availability is not
guaranteed all time because the choice of parent node
depends on the node decision to join or not the struc-
ture.
We also notice in Figure 5 that the time required
to obtain results from all sensors by using the tech-
nique of processing request is less than that of data
collection. With TinyDB, the data is regularly re-
ported and aggregated by a tree or a directed acyclic
graph from the nodes to an access point network. It
includes aggregation and filtering operations inside
the sensor network to maintain all routing informa-
tion. The parent nodes near the root put agreements
with their children on a time interval for the data lis-
tened to them. The whole process is repeated for each
period and query.
4.3.3 Impact of the MAC Layer Protocols
The goal of this test is to observe the Complete Cy-
cle Time (CCT) vs Duplicate Packets Sending (DPS)
with either a periodic data collection, or a query pro-
cessing system (TinyDB). We have varied the number
of sensors in each time.
We note in Table 5 that the number of transmission
packets with TinyDB is greater than that with peri-
SENSORNETS2015-4thInternationalConferenceonSensorNetworks
82
Table 5: Duplicate sends with periodic data collection and
TinyDB query processing.
Periodic data
collection
Query processing
(TinyDB)
Nb.nodes CCT(sec) DPS CCT(sec) DPS
5 47.430 0 2.203 0
10 53.955 0 2.434 0
20 54.157 0 4.766 21
30 62.082 0 14.060 102
40 73.302 0 25.441 121
50 106.609 0 435.43 174
60 147.357 2 705.02 294
70 179.478 1 112.261 308
80 266.357 15 142.114 405
90 457.570 78 233.737 420
100 570.100 72 313.832 633
odic data collection. The MAC-layer protocol can be
considered as one of the causes to explain these re-
sults. Tossim implements CSMA
1
based MAC proto-
cols by default. The basic principles of this protocol
are ”listen before talk and contention”. Each node lis-
tens on the channel and checks if it is clear before
each sending. If the channel is not clear, it waits for
a randomly chosen period of time called the backoff,
until the channel becomes clear. When the number
of sensor nodes increases, the nodes compete for the
channel and make the channel occupied. This com-
petition causes a transmission delays and a backoff
time accumulation, which increases the time needed
to collect data from all sensors. On the other hand,
each sender repeatedly sends its packet, during the
full wake-up, until it receives a link layer acknowl-
edgment from the receiver (sink node), meaning that
the receiver has correctly received the packet. In pe-
riodic data collection, the sky motes implement a ra-
dio duty cycling protocol, named ContikiMac using
a wake-up mechanism to listen for packet transmis-
sions from neighbors and to precise timing between
transmissions (Dunkels, 2011), which explains the
decrease in duplicate Packets Sending.
4.3.4 Impact of Choosing the Database on
Average Response Time
We use in this test three DBMS
2
: PostgreSQL,
MySQL, SQLite, to evaluate queries response time.
Table 6 shows that SQLite database gives the best
average response time for the SELECT queries, then
we find MySQL and finally PostgreSQL. For the IN-
SERT queries, PostgreSQL database has the less aver-
age response time compared to MySQL and SQLite.
SQLite takes a lot of time to insert data, because it
1
Carrier sense multiple access
2
DataBase Management System
does not have a central server to coordinate accesses.
It must close and reopen the database file, and in-
validate its cache, for each transaction. It does not
allow multiple concurrent writes in the same time.
PostgreSQL allows multiple transactions to proceed
with inserts concurrently. MySQL is better than Post-
greSQL for reading tables because it has a ”Query
Cache” to make queries faster. SQLite is faster than
MySQL because it uses a disk access to read and write
directly from the database files on disk.
Table 6: Query response time (ms).
Query type PostgreSQL MySQL SQLite
Insert queries 9.397 48.626 72.788
Select queries 0.992 0.690 0.225
4.3.5 Impact of Network Topologies on Data
Collection Time
We have tested 15 sender nodes and one sink node
with four network topologies. The result in Table 7
shows the time spent to collect data from all sensors
for the two technologies.
Table 7: Complete cycle time (ms).
Topologies Data collection Query processing
Star 62312 1832
Mesh 56002 1362
Grid 54121 2163
Tree 53515 2143
We note that the tree network topology for data
collection technology is more faster than the other
topologies. In fact, RPL forms a tree-like topology
rooted at the sink, reflecting the above results. Star,
mesh and grid topologies have not shown good re-
sults. RPL takes a time to to define DODAG networks
and to build path to the root. It can not fully exploit,
which affects the quality of paths. For the query pro-
cessing technology, the mesh topology gives the best
complete cycle time versus the other topologies. In
mesh topology, all nodes cooperate in the distribution
of data in the network. The same technology is used
by TinyDB. The nodes make an intelligent routing de-
cisions, where every node selects a parent based on
link quality to the base station.
5 CONCLUSION
In this work, we have studied and compared the tim-
ing properties of the data collection from a sensor
network using a static database versus an abstract
StudyofTimingPropertiesonDataCollectionandQueryProcessingTechniquesinWirelessSensorNetworks
83
in-network database, named TinyDB. We have eval-
uated temporal constraints such as data collection
time, database response time, and network conver-
gence time in both approaches. We have found, ac-
cording to the tests performed, that many factors can
affect the temporal constraints in a WSN. We have
determined that the network topology and the rout-
ing protocol together may play an important role on
data collection time. The convergence time also has
an impact on the process of data collection. We have
shown clearly the timing-response advantage of using
a TinyDB approach compared to accessing the data
stored in an external database. So we can conclude
that the great choice of the network topology and the
routing protocol with the right approach can improve
the temporal constraints in WSN. We plan to work in
this direction in a future works by taking into account
data temporal consistency.
REFERENCES
Amato, G., Chessa, S., and Vairo, C. (2010). Mad-wise:
A distributed stream management system for wireless
sensor networks. Softw. Pract. Exper., 40(5):431–451.
Bhattacharya, S. (2008). Achieving Application Quality of
Service in Resource-constrained Wireless Sensor Net-
works. PhD thesis, St. Louis, MO, USA.
Bonnet, P., Gehrke, J., and Seshadri, P. (2001). Towards
sensor database systems. In Proceedings of the In-
ternational Conference on Mobile Data Management,
pages 3–14, London, UK, UK. Springer-Verlag.
Choi, J. Y., Lee, J., Lee, K., Choi, S., Kwon, W. H., and
Park, H. S. Aggregation time control algorithm for
time constrained data delivery in wireless sensor net-
works. In Proceedings of the 63rd Vehicular Technol-
ogy Conference.
Dunkels, A. (2011). The ContikiMAC Radio Duty Cycling
Protocol. Technical Report T2011:13, Swedish Insti-
tute of Computer Science.
Fung, W. F., Sun, D., and Gehrke, J. (2002). Cougar the
network is the database. In International conference
on Management of data, pages 621–621, New York,
NY, USA. ACM.
Garc´ıa-Hernando, A., Mart´ınez-Ortega, J., L´opez-Navarro,
J., Prayati, A., and Redondo-L´opez, L. (2008). Prob-
lem Solving for Wireless Sensor Networks. Computer
Communications and Networks. Springer.
Gehrke, J. and Madden, S. (2004). Query processing in
sensor networks. IEEE Pervasive Computing, pages
46–55.
Hiromori, A., Uchiyama, A., Yamaguchi, H., and Hi-
gashino, T. (2012). Deadline-aware data collection
in csma/ca-based multi-sink wireless sensor networks.
In International Conference on Mobile Computing
and Ubiquitous Networking, pages 1–7.
Khoury, R., Dawborn, T., Gafurov, B., Pink, G., Tse, E.,
Tse, Q., Almi’Ani, K., Gaber, M. M., R¨ohm, U., and
Scholz, B. (2010). Corona: Energy-efficient multi-
query processing in wireless sensor networks. In DAS-
FAA (2), pages 416–419.
Laxaman, N., Goonatillake, M., and De Zoysa, K. (2010).
Tikiridb: Shared wireless sensor network database for
multi-user data access. CSSL.
Levis, P., Madden, S., Polastre, J., Szewczyk, R., White-
house, K., Woo, A., Gay, D., Hill, J., Welsh, M.,
Brewer, E., and Culler, D. (2005). TinyOS: An Open
Operating System for Wireless Sensor Networks. In
Ambient Intelligence, volume 35, pages 115–148.
Springer Berlin Heidelberg.
Madden, S., J. Franklin, M., M. Hellerstein, J., and Hong,
W. (2003). The design of an acquisitional query pro-
cessor for sensor networks. In Proceedings of the in-
ternational conference on Management of data.
¨
Osterlind, F., Dunkels, A., Eriksson, J., Finne, N., and
Voigt, T. (2006). Cross-level sensor network simu-
lation with COOJA. In The 31st Annual IEEE Confer-
ence on Local Computer Networks, Tampa, Florida,
USA, 14-16 November 2006, pages 641–648.
Sharaf, M. A., Beaver, J., Labrinidis, A., and Chrysanthis,
P. K. (2003). Tina: A scheme for temporal coherency-
aware in-network aggregation. In Proceedings of the
3rd ACM International Workshop on Data Engineer-
ing for Wireless and Mobile Access, MobiDe ’03,
pages 69–76, New York, NY, USA. ACM.
Suriyachai, P., Brown, J., and Roedig, U. (2010). Time-
critical data delivery in wireless sensor networks. In
Distributed Computing in Sensor Systems, pages 216–
229. Springer Berlin Heidelberg.
Vucinic, M., Tourancheau, B., and Duda, A. (2013). Perfor-
mance comparison of the rpl and loadng routing proto-
cols in a home automation scenario. In WCNC, pages
1974–1979.
Winter, T., Thubert, P., Brandt, A., Hui, J., Kelsey, R.,
Levis, P., Pister, K., Struik, R., Vasseur, J., and
Alexander, R. (2012). RPL: IPV6 Routing Protocol
for Low-Power and Lossy Networks.
SENSORNETS2015-4thInternationalConferenceonSensorNetworks
84