A Hybrid Strategy for Integrating Sensor Information

Koly Guilavogui, Laila Kjiri and Mounia Fredj

ENSIAS, Mohammed V-Souissi University, Avenue Mohammed Ben Abdellah Regragui, Rabat, Morocco

Keywords: Integration Systems, Data Warehouses, Multidimensional Databases, Mediation, Sensor Networks.

Abstract: The combination of sensor networks with databases has led to a large amount of real-time data to be

managed, and this trend will still increase in the next coming years. With this data explosion, current

integration systems have to adapt. One of the main challenges is the integration of information coming from

autonomously deployed sensor networks, with different geographical scales, but also with the combination

of such information with other sources, such as legacy systems. Two main approaches for integrating sensor

information are generally used: virtual and warehousing approaches. In the virtual approach, sensor devices

are considered as data sources and data are managed locally. In contrast, in the warehousing approach,

sensor data are stored in a central database and queries are performed on it. However, these solutions turn

out to be difficult to exploit in the current technology landscape. This paper focuses on the issue of

integrating multiple heterogeneous sensor information and puts forward a framework for decision making

process.

1 INTRODUCTION

During the last years, sensor networks have been an

area of intense investigation, from low-level

protocols to application level databases. They may

be wireless or not. Wireless Sensor Networks

(WSNs) are composed of small devices with low

processing power and data storage. In general,

sensor devices present limited energy and

communicate with each other by short-range radios

(Akyildiz et al., 2002). They produce raw data

continuously, thus providing data streams. The

elements of a stream can be numeric (e.g.

temperature), geospatial (e.g. GPS coordinates), or

multimedia (images, text, videos, and sounds). We

have then to deal with complex data.

The combination of WSNs with databases has

led to a large amount of real-time data to be

managed, and this trend will still increase in the next

coming years. With this data explosion, current

integration systems have to adapt. One of the

challenges is the integration of information coming

from autonomously deployed WSNs, with different

geographical scales, but also with the combination

of such information with other sources.

Nowadays, WSNs have been employed for

monitoring and controlling real world data in several

indoor and outdoor environments (Aggarwal et al.,

2013). They are gaining more and more attention by

both research communities and industries because of

their use in many applications (Akyildiz et al., 2002)

such as environmental monitoring (forest fire

prevention, flood control, etc.), monitoring traffic

(road, air, etc.), home automation, etc.

The great diversity of these sensors makes them

excellent tools for distributed sensing of phenomena,

processing and dissemination of information

collected to one or more observers (Aggarwal et al.,

2013).

This type of information is spatial, temporal,

thematic, and is inherently dynamic. This

complexity of information causes new needs in

terms of acquisition, structuring, integration,

analysis and visualization of this information for

decision support systems (Kimball and Caserta,

2004), (Favre et al., 2013).

It is undeniable that the use of sensors allows a

better monitoring of events that occur in the real

world. However, data produced by these devices

turn out to be difficult to exploit by conventional

approaches in data warehousing. Two main

approaches for integrating sensor information are

generally used: virtual and warehousing approaches.

In the virtual approach, sensor devices are

considered as data sources and data are managed

locally. In contrast, in the warehousing approach,

281

Guilavogui K., Kjiri L. and Fredj M..

A Hybrid Strategy for Integrating Sensor Information.

DOI: 10.5220/0004936302810286

In Proceedings of the 16th International Conference on Enterprise Information Systems (ICEIS-2014), pages 281-286

ISBN: 978-989-758-027-7

 2014 SCITEPRESS (Science and Technology Publications, Lda.)

sensor data are stored in a central database and

queries are performed on it. We note that the

approaches mentioned above are not mutually

exclusive, and can be combined in only one system

(this refers to the hybrid approach).

The design of an integration system requires not

only the choice of the approach (virtual,

warehousing or hybrid) but also the model of

integration. Basically, there are two models of

integration for specifying the correspondence or the

mapping between the data at the sources (local

schemas) and those in the global schema: the

Global-as-View (GaV) and the Local-as-View

(LaV). The other variants (hybrid approaches) are

just the combination of GaV and LaV (Halevy,

2001). It is exactly this mapping that will determine

how the queries posed to the system are answered

(Lenzerini, 2002).

In the GaV approach, the global schema is

modeled as a set of views over the schemas of the

sources. The major drawback of the GaV approach

is that there is a necessity to redefine the view of

the global schema every time a new source is

integrated. Therefore, it is an optimal choice when

the source schemas are not subject to frequent

changes. The LaV approach is the dual of the latter,

since the source schema is modeled as a set of views

on the global schema.

The major contribution of this paper is the

proposition of a hybrid approach which combines

the benefits of both virtual and data warehousing

approaches to efficiently and effectively integrate

data collected from WSNs and other sources. Our

framework follows the BGLaV (BYu Global-Local-

as-View) approach for the mapping between the

global schema and local schemas representing

sources of data. The BGLaV (Xu and Embley, 2004)

approach combines the best of LaV and GaV

approaches that uses source-to-target mappings

based on a predefined conceptual target schema. The

latter is specified ontologically and independently of

any of the sources.

Our objective is to provide an information

integration system that uses not only any type of

data sources (relational, semi-structured, ontologies,

etc.) but also ensures the freshness of data and

provides more flexibility with respect to adding or

deleting a new information source. We also aim at

building and composing multidimensional databases

(data cubes) on-the-fly using our integration system.

Existing approaches for integrating sensor

information have some drawbacks and do not meet

aforementioned requirements.

The remainder of this paper is organized as

follows. Section 2 reviews the state of the art

concerning the integration of heterogeneous sensor

information with other sources. Section 3 puts

forward the proposed architecture for the hybrid

strategy. Section 4 concludes with a perspective of

our on-going work.

2 RELATED WORK

As WSNs can be considered as autonomous,

distributed and heterogeneous sources, data

integration was needed in order to give the user an

integrated view of information sources.

2.1 Data Integration Challenges

Data integration is vital in large organizations in

order to deal with a research problem or

environmental issues (e.g. forest fire, flooding, water

quality, etc.). These organizations generally own a

multitude of data sources, where data sets are being

produced independently. In practice, sensor schemas

typically refer to relational database schemas, semi-

structured information or ontologies (Konstantinou,

2012).

The main challenge for an integration system is

to solve different conflicts among information

sources and to represent them in a single coherent

schema (Ziegler and Dittrich, 2004). These conflicts

may be encountered at schema and instance level.

Beyond the issues of structural integration of

heterogeneous information, the challenge is to

identify conflicts among concepts in different

sources that are semantically related, and then to

propose a resolution of those conflicts. This data

integration process is quite complex, since the

integration system has also to take into account

issues related to autonomy, distribution, dynamicity

and volume of data.

In the next section, we review the virtual,

warehousing and hybrid approaches for sensor

information integration and we make a comparative

analysis.

2.2 Virtual Approach

In this approach in Figure 1 proposed by

(Wiederhold, 1992), the integration system is based

on a mediator with a global schema and wrappers.

Users’ queries are expressed on the global schema.

Two steps are required: the mediator accepts a query

and determines the appropriate set of information

sources to answer the query, and then generates the

ICEIS2014-16thInternationalConferenceonEnterpriseInformationSystems

282

appropriate sub queries or commands for each

information source. Wrappers perform translation,

filtering and data collection from local sources.

Then, the mediator obtains results from the

information sources via wrappers, merges

information and returns the final answer to the user

or application.

Figure 1: Virtual approach of information integration

(Wiederhold, 1992).

This approach refers to virtual, mediation, on-

demand or on-the-fly approach, since information is

extracted from the sources only when queries are

posed to the system. The Cougar system (Yao and

Gehrke, 2002) at Cornell University is the first step

toward sensor data management using the virtual

approach. The work of (Ibrahim et al., 2005; Stocks

et al., 2009) both also developed a mediation system

using sensors. The work of Casola et al. (2009)

proposed SeNsIM (Sensor Networks Integration and

Management). SeNsIM is a framework that enables

the integration of heterogeneous sensor systems

using XML as modeling language.

2.3 Warehousing Approach

In this approach in Figure 2, a new physical database

is created to integrate and import data from several

sources, usually in one direction only (Widom,

1995; Inmon, 1996; Kimball and Caserta, 2004).

In the warehousing approach, the data model is

multidimensional and queries are generally complex

using sophisticated means of navigation in data

across different dimensions with OLAP (On Line

Analytical Processing) operators (Codd and Salley,

1993).

In Figure 2, wrappers do the extraction and

some transformation tasks while the integrator

performs the final transformation and loading

process. Queries are directly evaluated on the data

warehouse without accessing the original

information sources.

In this paper, we make a difference between the

materialized and the warehousing approaches,

although the two approaches are almost similar. The

materialized approach generally refers to data

warehousing in some studies (Zhou et al., 1995;

Shokoh, 2010). In fact, data warehouses (DW) use

materialize views, but the materialized approach

does not all include the requirements for data

warehousing. A materialized view is thus like a

cache - a copy of data that can be accessed quickly

(Gupta and Mumick, 1995). This approach is not

intended to provide analytical processing. Opposite

to DW, the historical data is not kept for long time

storage and the new version of data replaces the last

one.

Figure 2: Warehousing approach of information

integration (Widom, 1995).

Recent works (Shah et al., 2009; Ahmed et al., 2010;

Li et al., 2013; Gökçe and Gökçe, 2014) used the

warehousing approach for reducing the energy

consumption in buildings equipped with sensors.

(Da Costa and Cugnasca, 2010) also used a

warehousing approach to manage sensor data for

animal monitoring (pollinators). (Mathieu, 2011)

proposed a web service based architecture based on

OGC (Open Geospatial Consortium) standards for

near real-time integration of sensor data streams into

a geo-analytical data warehouse.

2.4 Hybrid Approach

This approach is usually used to overcome the

drawbacks of the approaches mentioned above. In

this work, we consider the hybrid approach to be

divided in two: (a) the hybrid approach that

combines virtual and warehousing approaches; and

User

Integrator

Data warehouse

Wrapper

WSN

Wrapper

WSN

Wrapper

Legacy source

Wrapper

WSN

…

User

Mediator

Wrappe

WSN

Wrapper

WSN

Wrapper

WSN

Wra

source

AHybridStrategyforIntegratingSensorInformation

283

(b) the hybrid approach that combines virtual and

materialized approaches. The work of (Grosky et al.,

2007; Huang and Javed, 2008; Roantree, 2009)

followed the hybrid approach (virtual and

materialized) for data integration. In the next

section, we make a comparative analysis of the

different approaches we consider in this work.

2.5 Comparative Analysis

As summarized in Table 1, there is no work that

handles all the type of data sources that can be found

in WSNs environment. However, all the authors

used relational data sources in their respective work.

One can see from Table 1 that the virtual

approach allows flexibility and freshness of data but

avoids data replication. It is not intended to OLAP

manipulation but for information retrieval.

Generally, the virtual approach is preferable to the

warehousing approach in the following cases

(Convey et al., 2001): (a) the number of data sources

in the integration system is very large, scalable, and

sources are likely to be updated frequently, and (b)

there is no way to predict the types of queries that

the user will make.

The warehousing approach allows data

replication and it is well suited for OLAP of

historical data, but the major drawbacks of this

approach are freshness of data and flexibility when

adding or deleting new information sources.

However, the hybrid approach is suited in case

some data sources of underlying sources are

frequently changing and other may change less

frequently (Shokoh, 2010). Research using a hybrid

approach is not widespread. The existing works that

used this hybrid approach generally built their

integration system for the purpose of information

retrieval and not for OLAP. Our work is different,

since our framework not only uses virtual and data

warehousing approaches but also data sources could

be from any types (relational, semi-structured,

unstructured, geospatial or ontologies).

To the best of our knowledge, the hybrid

approach (virtual and data warehousing) has not

been used for managing sensor information

combined with other sources. In the next section, our

Table 1: Comparative analysis of the three approaches.

Approaches

Criteria

Virtual Warehousing Hybrid

(Yao and

Gehrke, 2002)

(Ibrahim and

al., 2005)

(Casola et al.,

2009)

(Stocks et al.,

2009)

(Shah et al.,

2009)

(Ahmed et al.,

2010)

(Da Costa and

Cugnasca,

2010

)

(Mathieu,

2011)

(Li et al., 2013)

(Gökçe and

Gökçe, 2014)

(Grosky et al.,

2007)

(Huang and

Javed, 2008)

(Roantree,

2009)

Type of data sources

Relational (R)

/Object-

relational

(O-R)

+ + + + + +

+ + + + + + +

Semi-

structured

(XML)

- - + - - - + + - + + + +

Ontologies

(RDF/OWL)

- - - - - - - - - - - + -

Unstructured

data (CSV,

videos, etc.)

- - - + - +

- - + +

- +

Geospatial

data type

- - - + - - - + - + + - +

Flexibility

+ + + + - - - - - - + + +

Freshness

+ + + + - - - + - - + + +

Data replication

- - - - + + + + + + + + +

Information

retrieval

+ + + + - - - - - - + + +

OLAP manipulation

- - - - + + + + + + - - -

Modeling language

ADT

objects

R XML XML OWL R R R R R - RDF O-R

ADT: Abstract Data Type; R: Relational model; OWL: Ontology Web Language; O-R: Object-Relational; RDF: Resource Description

Framework; XML: eXtensible Markup Language

ICEIS2014-16thInternationalConferenceonEnterpriseInformationSystems

284

proposed architecture is presented.

3 OUR PROPROSED

ARCHITECTURE

We put forward a hybrid strategy named MEDWare

(Mediation and Data Warehousing framework). We

describe the main components of our framework

which is divided into four layers (see Figure 3):

(a) Sensor network information sources: this

layer consists of heterogeneous WSNs and other

sources (e.g legacy systems).The information from

the sensor nodes are gathered and can be accessed

through a standard sensor gateway (a device uses for

interfacing between the WSNs and a computer).

(d)

(c)

(b)

(a)

Figure 3: Our architecture based on a hybrid strategy.

(b) Mediation layer: consists of the mediator

that maintains the global schema and mappings

between the global and source schemas. The

mediator also ensures query processing and solves

semantic heterogeneity. Wrappers are also in that

layer. They build a procedure to extract data from

the sources and deal with syntactic heterogeneity.

composed of the data warehouse and

multidimensional databases (OLAP data cubes) that

can be built on-the-fly.

(d) Application layer: consists of different client

applications used by the user to submit queries to the

system for OLAP of sensor data, searching sensor

information (e.g meta-data), or to receive

notifications (e.g alerts).

The main advantages of this architecture are the

following: (1) the development of a flexible decision

support system (DSS) where new sensor networks or

other sources can be easily added or removed from

the system, (2) the freshness of data in the DSS, (3)

the rapid building and composition of data cubes on-

the-fly for better decision making, and (4) the ability

to search information in heterogeneous sources.

The combination of the virtual and data

warehousing approaches does not occur without

problems. One can imagine the impact of such a

combination on the performance analysis. Indeed,

the proposed architecture generates many challenges

to be addressed for future contributions such as:

- The query response time: when the number of

supported users that interact with the system

grows, it is necessary to deal with the issue of

minimizing the query response time.

- Description of sensor data: there is a lack of

adding semantics to sensor data in order to deal

with semantic heterogeneity. Most of the time,

the semantics of data is implicit.

4 CONCLUSIONS

In this paper, we proposed an architecture that

integrates sensor information from multiple

heterogeneous sensor networks and other sources for

decision support systems.

Currently, we are working on formal description

of the system, example scenarios and semantic data

integration issue. Since then, many ontologies have

been developed for describing sensor network

information in the context of the semantic sensor

web, sensor ontology integration is needed. The

implementation is ongoing in order to validate the

proposed architecture.

REFERENCES

Aggarwal, C. C., Ashish, N., Sheth, A. 2013. The internet

of things: a survey from the data-centric perspective,

in: Managing and Mining Sensor Data. Springer, pp.

383–428.

Building the context analysis

User’s query

Searching, OLAP, Data mining, Alerting, etc.

WSN

Wrapper

Gateway

WSN

Mediator

(Query Reformulation, Optimization, Query Execution)

Wrapper

Legacy source

Cube

Data extracted

Data warehouse

AHybridStrategyforIntegratingSensorInformation

285

Ahmed, A., Ploennigs, J., Menzel, K., Cahill, B., 2010.

Multi-dimensional building performance data

management for continuous commissioning. Advanced

Engineering Informatics 24(4), 466–475.

Akyildiz, I. F., Su, W., Sankarasubramaniam, Y., and

Cayirci, E. 2002. A survey on sensor

networks. Communications magazine, IEEE, 40(8),

pp. 102-114.

Casola, V., Gaglione, A., Mazzeo, A. 2009. A reference

architecture for sensor networks integration and

management, in: GeoSensor Networks. Springer, pp.

158–168.

Codd, E. F., Codd, S. B., and Salley, C. T., 1993.

Providing OLAP (on-line analytical processing) to

user-analysts: An IT mandate. Codd and Date, 32.

Convey, C., Karpenko, O., and Tatbul, N., 2001.

Data integration services. http://www.cs.brown.

edu/people/atbul/cs227/chapter.pdf.

Da Costa, R., and Cugnasca, C. E. 2010. Use of data

warehouse to manage data from wireless sensors

networks that monitor pollinators. In 11th

International Conference on Mobile Data

Management (MDM), IEEE Computer Society,

Missouri, USA. pp.402-406.

Favre, C., Bentayeb, F., Boussaid, O., Darmont, J., Gavin,

G., Harbi, N., ... and Loudcher, S., 2013. Les entrepôts

de données pour les nuls... ou pas!. 2ème Atelier Aide

à la Décision à tous les Etages (EGC/AIDE 13),

Toulouse, France.

Gökçe, H. U., Gökçe, K. U., 2014. Multi-dimensional

energy monitoring, analysis and optimization system

for energy efficient building operations. Sustainable

Cities and Society, 10, pp. 161–173.

Grosky, W. I., Kansal, A., Nath, S., Liu, J., Zhao, F., 2007.

Senseweb: An infrastructure for shared sensing.

Multimedia, IEEE, 14, pp. 8–13.

Gupta, A., and Mumick, I. S., 1995. Maintenance of

materialized views: Problems, techniques, and

applications. IEEE Data Eng. Bull., 18(2), pp. 3-18.

Halevy, A.Y. 2001. Answering queries using views: A

survey. The VLDB Journal, 10(4), pp. 270-294.

Huang, V., Javed, M.K., 2008. Semantic sensor

information description and processing, in: Sensor

Technologies and Applications. SENSORCOMM’08.

pp. 456–461.

Ibrahim, I. K., Kronsteiner, R., and Kotsis, G., 2005. A

semantic solution for data integration in mixed sensor

networks. Computer Communications, 28(13), pp.

1564-1574.

Inmon, W. H. 1996. Building the data warehouse. Wiley

Publishing, Inc.

Kimball, R., and Caserta, J. 2004. The data warehouse

ETL toolkit: practical techniques for extracting,

cleaning, conforming, and delivering data. Wiley

Publishing, Inc.

Konstantinou, N., 2012. Converting raw sensor data to

semantic web triples: a survey of implementation

options. Journal of Sensors, Wireless Communications

and Control 2.1: 44-52.

Lenzerini, M., 2002. Data integration: a theoretical

perspective. In Proceedings of the twenty-first ACM

SIGMOD-SIGACT-SIGART symposium on Principles

of database systems, ACM, pp. 233-246.

Li, Y., Wang, L., Ji, L., Liao, C., 2013. A data warehouse

architecture supporting energy management of

intelligent electricity system, in: Proceedings of the

International Conference on Computer Science

and Electronics Engineering. Atlantis Press, Paris,

France.

Mathieu, J. 2011. Intégration de données temps-réel

issues de capteurs dans un entrepôt de données géo-

décisionnels. Thesis (MSc), Université de Laval.

Canada.

Roantree, M., 2009. A hybrid storage model for web

information systems, in: Proceedings of the 6th

International Workshop on Web Information Systems

Modeling. Amsterdam, The Netherlands.

Shah, N., Tsai, C. F., Marinov, M., Cooper, J., Vitliemov,

P., and Chao, K. M., 2009. Ontological on-line

analytical processing for integrating energy sensor

data. IETE Technical Review, 26(5), pp. 375.

Shokoh, K. 2010. A hybrid approach to data integration.

Thesis (PhD). Université Joseph Fourier-Grenoble I.

France.

Stocks, K. I., Condit, C., Qian, X., Brewin, P. E., and

Gupta, A., 2009. Bringing together an ocean of

information: an extensible data integration framework

for biological oceanography. Deep Sea Research Part

II: Topical Studies in Oceanography, 56(19), pp.

1804-1811.

Widom, J. 1995. Research problems in data warehousing.

In Proceedings of the fourth international conference

on Information and knowledge management, USA,

ACM, pp. 25-30.

Wiederhold, G. 1992. Mediators in the architecture of

future information systems, IEEE Computer, 25(3),

pp. 38–49.

Xu, L., and Embley, D. W., 2004. Combining the best of

Global-as-View and Local-as-View for data

integration. In ISTA, 48, pp. 123-136.

Yao, Y., and Gehrke, J., 2002. The cougar approach to in-

network query processing in sensor networks. ACM

Sigmod Record, 31(3), pp. 9-18.

Zhou, G., Hull, R., King, R., and Franchitti, J. C. 1995.

Data integration and warehousing using H2O. IEEE

Data Eng. Bull., 18(2), pp. 29-40.

Ziegler, K., and Dittrich, R. 2004. Three decades of Data

Integration – All Problems solved? Building the

Information Society, Springer.

ICEIS2014-16thInternationalConferenceonEnterpriseInformationSystems

286