A QUALITY OF CONTEXT DRIVEN APPROACH FOR THE
SELECTION OF CONTEXT SERVICES
Elarbi Badidi
1
and Larbi Esmahi
2
1
College of Information Technology, United Arab Emirates University, PO.Box. 17551, Al-Ain, United Arab Emirates
2
School for Computing & Information Systems, Athabasca University
1 University Drive Athabasca, Alberta, T9S 3A3, Canada
Keywords: Mobile Workers, Publish-subscribe, Quality-of-context, Context-aware Web Services, Service Selection.
Abstract: As wireless networks and mobile devices are becoming ubiquitous, mobile users are increasingly requiring
access to application services that can adapt to their context as they move to new locations, for example, in
their corporate or partners' networks. The quality-of-context information (QoC) used by these application
services is a determinant factor in the adaptation process. As application services typically receive context
data from several context services, the selection of suitable context services is of paramount importance in
providing mobile users with tailored services. In this paper, we describe our proposed framework for
context management and our proposed QoC-based algorithm for the selection of context services. The
algorithm takes into account the QoC requirements of application services for each context information to
which they subscribe with the Context Broker, on which the framework is relying for context management.
1 INTRODUCTION
As wireless networks and mobile devices are
becoming ubiquitous, there is an unprecedented rise
in the number of mobile workers who are using a
variety of modern handheld devices such as PDAs
and SmartPhones to consume online services.
With this proliferation of mobile devices, wireless
business applications, i.e. messaging and voice
services, and healthcare services are more and more
developed and deployed using Web services. Thus,
using the Service Oriented Computing paradigm in
mobile environments considerably enlarges the
range of accessible business applications and
enables delivering integrated services across
wireless networks.
Mobile workers are increasingly requiring
retaining access to services that are similar to their
corporate services as they move to new locations in
a very straightforward manner without having to
configure their working environment explicitly.
Therefore, business services should be context-
aware to deal with the changing environment of the
user. Several definitions of the notion of context
have been provided in the literature. According to
Dey’s definition (Dey (2001)), the amount of
information that can be categorized as context
information is extremely broad. Location and time
are the most widely used context parameters by
applications. Wireless devices such as
environmental sensors, radio frequency
identification (RFID) tags send raw context
information such as location, presence and other
status information across the network. Specialized
services, that we call context services, capture, store,
analyze and aggregate data to provide high-level
context information to application services as
needed. Indicators that can essentially be measured
and captured include temperature, humidity,
pressure, whether the mobile worker is in motion,
and many more.
Being context-aware allows application services
to make inferences about the current situation of the
user using context information obtained from
various sources such as surrounding sensors, and
GPS receivers, and hence adapt their behavior to the
prevailing context of the user. By fusing acquired
data and inferring on it, these applications can
deduce, for instance, the position or the current user
activity such as “user is in a shopping mall” or “user
is in a meeting”. Such Context-aware application
services are being deployed in various industries
89
Badidi E. and Esmahi L. (2010).
A QUALITY OF CONTEXT DRIVEN APPROACH FOR THE SELECTION OF CONTEXT SERVICES.
In Proceedings of the International Conference on e-Business, pages 89-94
DOI: 10.5220/0003031200890094
Copyright
c
SciTePress
such as healthcare, airline transportation,
manufacturing, and retail. In the last few years,
Context-aware Web services have emerged as a
promising technology for building innovative and
interoperable context-aware applications (Truong et
al. (2009)).
One of the main problems being faced in the area
of ubiquitous computing is handling and distributing
context efficiently to enhance personalized service
delivery to mobile users. The context sources as well
as the consumer services of context information are
very often physically distributed. For instance, the
context sources providing information about the
current temperature may be far from the applications
that need to adapt their services to the prevailing
weather conditions. Furthermore, it is likely that
these context sources provide the same context
information but with different QoC (Buchholz et al.
(2003); Kamran et al. (2008)). Context-awareness
raises new challenges like aggregation of context
information in a structured format, discovery and
selection of suitable context sources.
In this paper, we describe our proposed
framework for context management and our
proposed QoC-based algorithm for the selection of
context services by application services, so that they
can adapt their services to the user context. The
algorithm takes into account the QoC requirements
of the application services for each context
information to which they subscribe with the
Context Broker. The framework is relying on the
Context Broker for context management. Selection
of context-aware application services by users or
intermediary entities is not the subject of this work
as it has been investigated by several research works
(Yu et al. (2009); Kirsch-Pinheiro et al. (2008)). To
the best of our knowledge, selection of context
services has not been investigated before.
The Context Broker implements a topic-based
publish/subscribe model in which application
services subscribe to context information they are
interested in, and context services publish their
context information. Context Brokers have been
used also in the following works (Chen et al. (2003);
Bonino da Silva Santos et al. (2007)).
2 BACKGROUND
2.1 Quality-of-Context
Context information is characterized by some
properties referred in literature as quality-of-context
(QoC) indicators. Buchholz et al. (Buchholz et al.
(2003)) have defined the QoC as: “Quality of
Context (QoC) is any information that describes the
quality of information that is used as context
information. Thus, QoC refers to information and
not to the process nor the hardware component that
possibly provide the information.”
Buchholz et al. (2003)) and Kamran et al. (2008)
have identified the following QoC indicators:
precision, freshness, temporal resolution, spatial
resolution, and probability of correctness.
Precision
. This indicator represents the granularity
with which context information describes a real
world situation.
Freshness
. The time that elapses between the
determination of context information and its delivery
to a requester.
Spatial Resolution
. The precision with which the
physical area, to which an instance of context
information is applicable, is expressed.
Temporal Resolution
. The period of time during
which a single instance of context information is
applicable.
Probability of Correctness
. This indicator represents
the probability that a piece of context information is
correct.
Several competing context services may provide the
same context information (Buchholz et al. (2003)).
Therefore, potential context consumers should be
able to select context services on the basis of the
QoC they can assure.
2.2 Context Services
A context service typically provides infrastructure
support for collection, management, and
dissemination of context information concerning a
number of subjects. Subjects may be users, objects
such as handheld devices and equipments, or the
environment of users. The context service acquires
context information from various context sources.
Sources are usually third parties that collect and
provide context information. For example, consider
the “temperature” at the current location of the
mobile user. This information may be obtained
directly from the mobile device of the user. It can
also be obtained from a local weather station.
Alternatively, it may be obtained from weather TV
channels providing weather information nation-
wide.
Several research works have investigated the
design and the implementation of context services.
Shmidt et al. designed and implemented a generic
context service with a modular architecture that
ICE-B 2010 - International Conference on e-Business
90
Figure 1: Framework Components.
allows for context collection, discovery and
monitoring (Shmidt et al. (2009)). This context
service provides a Web service interface that allows
its integration in heterogeneous environments. The
implementation uses OWL to describe context
information and SPARQL to query and monitor
context information.
Lei et al. described the design issues and the
implementation of a middleware infrastructure for
context collection and dissemination (Lei et al.
(2002)). They realize this middleware infrastructure
as a context service. To allow for wide deployment
of the context service, this work has addressed the
following issues: Extensibility of the context service
architecture by supporting heterogeneous context
sources, integrated support for privacy, and quality
of context information support. Coronato et al.
proposed a semantic context service that relies on
semantic Web technologies to support smart offices
(Coronato et al. (2006)). It uses ontologies and rules
to infer high-level context information, such as
lighting and sound level, from low-level raw
information acquired from context sources.
3 FRAMEWORK OVERVIEW
As illustrated in Figure 1, our proposed framework
is relying on a Domain Broker, which mediates
between mobile users and application services at the
visited site. The Domain Broker components are the
QoS Broker and the Context Broker. Another key
component of the framework is the Policy Manager,
which is responsible for managing and maintaining
authentication and authorization policies, as well as
polices for monitoring services and their quality of
service. The Domain Broker components collaborate
to deliver personalized context-aware services to
mobile users with various devices across interacting
sites. The QoS Broker is in charge of managing the
QoS of Application Web services and submitting the
mobile user requests to suitable ones. The Context
Broker is in charge of managing context information
and user profile and preferences. QoS Management
operations (QoS specification, monitoring, service
level agreement (SLA) negotiation) performed by
the Qos Broker components are described in our
previous work (Badidi et al. (2009)). This paper
extends the proposed architecture by considering the
context dimension so that applications services are
context-aware.
When applications are context-aware, they can
adapt their behavior and offer to the user
contextually relevant information. Knowledge of the
user context allows anticipating the user service and
information needs. Context management is achieved
in our framework by using a Context Broker and by
adopting a topic-based Publish/subscribe messaging
model, a one-to-many pattern of asynchronous
message distribution based on registration of
interest. Publishers label each message with the
name of a topic (“publish”) rather than addressing it
directly to subscribers. The message system then
sends the message to all eligible recipients that
expressed their interest in receiving messages on that
topic (“subscribe”). The publish/subscribe model is a
loosely coupled architecture in which senders often
do not need to know who their potential subscribers
A QUALITY OF CONTEXT DRIVEN APPROACH FOR THE SELECTION OF CONTEXT SERVICES
91
are, and the subscribers do not need to know who
generates the information.
Figure 2: Topic-based publish/subscribe system.
It is increasingly being used in a service oriented
architecture context. A Web service disseminates
information to a number of other Web services,
without the need to have prior knowledge of these
other services.
The Context Broker is a mediator Web service
that decouples context consumers from context
services. It is in charge of handling subscriptions of
application Web services in which they express their
interest to receive context information, and
registrations of context services. Once context
services are registered with the Context Broker, they
publish context information and application Web
services are then notified by the Context Broker
about the new context information. Figure 2
illustrates our topic-based publish/subscribe system
in which context services are the publishers and the
application services are the subscribers. Context
information -- such as location, temperature, and
user activity -- represents the topics of the system.
4 QoC-BASED CONTEXT
SERVICES SELECTION
As we have mentioned earlier, context information
may be delivered with different QoC by various
context services. Therefore, the Context Broker is in
charge of selecting appropriate context services to
deliver context information to which application
services subscribe. Context information may be
delivered to the same application service by several
context services. Each one may deliver a piece of
context information (a topic) that the application
service requires to adapt its behavior to the current
context.
In the following, we describe the context services
selection algorithm for a given application service.
Let
,
,…,
be the list of context
information (topics) to which an application service
has subscribed by showing its interest in receiving
such context information.
Let
,
,…,
be the list of context
services registered with the Context Broker. Two
context services may provide different context
information; each one specializes in offering
particular context information. One service, for
example, may offer location information while
another service may offer only temperature
information, and a third one may offer both of them.
These services typically provide context
information with different QoC. We assume that
QoC indicators are in normalized form with values
between 0 and 1. A value of 1 means highest quality
and 0 means lowest quality. For example for the
freshness quality indicator, 1 means that context
sources have sensed the information in the last
minute, and 0 means that they have sensed it in the
last 10 minutes.
When subscribing to context information, an
application service specifies the min values of the
normalized QoC indicators that it can tolerate. For
instance, the application service may subscribe to
the location information may require a min value of
75% for the freshness indicator and 95% for the
probability of correctness indicator. Let
,
,…,
be the list of QoC indicators
(parameters) considered in the system.
The minimum quality requirements that the
application service tolerates for a given context
information (topic)
,with 1,are
expressed by the following vector:
,
,
,
,…,
,
0
,
1, with 1 and is the
cardinality of .
Therefore, the whole quality-of-context
requirements of the application service for all its
subscribed topics and all QoC indicators considered
in the system can be expressed by the following
matrix:
………..
…
…
,
,
,
,
,
,
,
,
,
The goal of the selection algorithm is to find for
each topic
, to which the application service
subscribed, a suitable context service from the set
ICE-B 2010 - International Conference on e-Business
92
that can satisfy the minimum quality requirements of
the application service.
The QoC offer of a context service CS
is
expressed by the following matrix:
………..…
…
…
,
,
,
,
,
,
,
,
,
CS
is suitable for a topic
if the following
condition is satified:
0
,
,
1 for 1 and 1 (1)
We define the distance of the QoC offer of CS
from the application service required QoC for each
quality indicator as:
,
,
,
for 1 and 1
Therefore, we can consider the distance matrix
for the QoC offer CS
.
………..…
…
…
,
,
,
,
,
,
,
,
,
Using the distance matrix, we can say that the
application service requirement can be satisfied for a
given topic
by the context service CS
if the
corresponding row in the above matrix has values
greater than or equal to zero. Therefore, we can
discard from that matrix the rows having negative
values. We call the resulting matrix
(all rows
have positive values).
The Euclidian distance of CS
offer from the
application service QoC requirement for topic
is:
∑
,
with 1 (2)
The highest value of
corresponds to the best QoC
offer that can fulfill the QoC requirements of the
application service for the topic
.
The most suitable context service for topic
,
that we call here
, will be the one that
maximizes the above Euclidian distance, that is :
max
∑
,
(3)
max
∑
,
,
(4)
In (4) we have assumed that the application service
gives the same weight to all QoC indicators. This is
not always the case as the application service may
set relative weights for the QoC indicators. The
application service may even set weights for each
topic to which it subscribed. For example, for the
location topic, more weight may be given to the
spatial resolution indicator than to the probability of
correctness indicator. For the time of the day topic,
more weight may be given, for example, to the
precision indicator than to the other QoC indicators.
Therefore the weights for a given topic
, can be
expressed by the following vector:
,
,
,
,…,
,
Where 0
,
1 and
∑
,
1
Given these weights, the most suitable context
service for the topic
,
, will be the one that
satisfies the condition (1) and which maximizes the
sum of quality offers for all QoC indicators:
max
∑
,
,
(5)
If no context service satisfies the application service
QoC requirements for a given topic, then the
Context Broker may ask the application service to
lower its QoC expectations.
The key idea of the above QoC-based context
service selection algorithm is to find the most
suitable context service with regard to the QoC
requirements of a given application service for each
context information (topic) to which the application
service has subscribed.
5 PROOF OF CONCEPT
As a proof of concept, we describe in this section a
scenario of how the QoC selection algorithm works.
Assume that the topics to which an application
service has subscribed are in order of location, time
of day, activity, and temperature.
Assume that the QoC indicators considered in the
system are in order of freshness, probability of
correctness, temporal resolution, and spatial
resolution. The normalized minimum QoC
requirements of the application service are given in
Table 1.
Assume that four context services have
registered with the context broker, CS
1
, CS
2
,
CS
3
,
and
CS
4
. In the following we consider only the location;
the same process applies to the other topics. Table 2
describes The QoC offer of the four Context services
for the location context information (topic).
A QUALITY OF CONTEXT DRIVEN APPROACH FOR THE SELECTION OF CONTEXT SERVICES
93
Table 1: Minimum QoC requirements of the application
service.
freshness Probability
correctness
Temporal
resolution
Spatial
resolution
location 0.95 0.75 0.65 0.75
time 0.80 0.85 0.80 0.65
activity 0.65 0.75 0.65 0.50
temperature 0.85 0.90 0.50 0.90
Table 2: QoC offers of four Context service for the
location topic.
freshness Probability
correctness
Temporal
resolution
Spatial
resolution
CS
1
0.97 0.80 0.75 0.85
CS
2
0.80 0.70 0.80 0.65
CS
3
0.98 0.78 0.65 0.80
CS
4
0.95 0.80 0.70 0.80
In this scenario, we consider that the application
service assigns the same weight to all QoC
indicators. By computing the distance matrix for
QoC offers for the location topic and the Euclidian
distance for each context service, we get the
following ranking of the context services from
highest offer to lowest:
CS
1
(0.1513), CS
4
(0.0866), CS
3
(0.0.0656)
The QoC offer of CS
2
does not meet the minimum
QoC requirements of the application service.
6 CONCLUSIONS
Context services are increasingly used as an
intermediary between context-aware application
services and context sources. They provide
infrastructure support for collection, management,
and dissemination of context information concerning
a number of subjects. The adaptation of services to
the context of users requires the acquisition of high-
quality information from context services.
Therefore, the selection of suitable context services
is becoming a pressing issue. In this paper, we have
presented our proposed framework for context
management and our proposed QoC-based algorithm
for the selection of context services. The algorithm
takes into account the QoC requirements of the
application services for each context information to
which they subscribe with the Context Broker, on
which the framework is relying for context
management.
REFERENCES
Badidi, E., Esmahi, L., Serhani, M. A., and Ikbal, T.,
2009. A QoS-Broker Based Framework for
Personalized Web Services Provisioning to Mobile
Users, In Proceedings of the International
Conference on Wireless Information Networks and
Systems (WINSYS'09), Milan, Italy, July 2009, 147-
152.
Bonino da Silva Santos, L. O., van Wijnen, R. P., and
Vink, P., 2007. A service-oriented middleware for
context-aware applications. In Proc. of the 5th
international Workshop on Middleware For
Pervasive and Ad-Hoc Computing at Middleware
2007, pp.37-42.
Buchholz, T., Kpper, A., Schiffers, M., 2003. Quality of
context: What it is and why we need it. In Proc. of
the 10th International Workshop of the HP OpenView
University association (HPOVUA).
Chen, H., Finin, T., and Joshi, A., 2003. An ontology for
context-aware pervasive computing environments.
Knowledge Engineering Review 18: pp.197-207
Cambridge University Press.
Coronato, A., De Pietro, G., and Esposito, M., 2006. A
Semantic Context Service for Smart Offices. In Proc.
of the 2006 international Conference on Hybrid
information Technology - Volume 02, pp.391-399.
Dey, A.K., 2001. Understanding and Using Context.
Journal of Pervasive and Ubiquitous Computing,
5(1), pp. 4–7, 2001.
Kirsch-Pinheiro, M., Vanrompay, Y., and Berbers, Y.,
2008. Context-Aware Service Selection Using Graph
Matching. In Proc. of the 2nd Workshop on Non
Functional Properties and Service Level Agreements
in Service Oriented Computing at ECOWS 2008.
Lei, H., Sow, D. M., Davis, J. S., Banavar, G., and Ebling,
M.R., 2002. The design and applications of a context
service. SIGMOBILE Mob. Comput. Commun.
Rev. 6, 4 (Oct. 2002), pp.45-55.
Schmidt, H., Flerlage, F., Hauck, F.J., 2009. A generic
context service for ubiquitous environments. In Proc.
of the IEEE International Conference on Pervasive
Computing and Communications (PERCOM 2009),
pp.1-6.
Sheikh, K., Wegdam, M., and van Sinderen, M., 2008.
Quality-of-Context and its use for Protecting Privacy
in Context Aware Systems. Journal of software, 3(3),
pp.83-93, March 2008
Truong, H.L., and Dustdar, S., 2009. A Survey on
Context-aware Web Service Systems. International
Journal of Web Information Systems, 5(1), pp.5-31,
Emerald, 2009.
Yu, H.Q., and Reiff-Marganiec, S., 2009. Automated
Context-aware Service Selection for Collaborative
Systems. In Proc. van Eck, J. Gordijn, R. Wieringa
(eds.): Advanced Information Systems Engineering
2009 (CAiSE'09), LNCS 5565, pp. 261-274.
ICE-B 2010 - International Conference on e-Business
94
