![](bg2.png)
individual service elements at run time just like the
dynamic service composition scenario. Furthermore,
the game assigns ‘scores’ to players on the basis of
their performance. A group of volunteers are re-
quested to play the game and their respective scores
are noted. Subsequently, the game is played with
the service selection decisions being made using our
composition model. The score of our model is com-
pared with those of the volunteers. If the model con-
sistently outperforms the players, the model’s efficacy
is validated.
The remainder of this paper is organized as fol-
lows: Section 2 is a discussion on other techniques of
validation for dynamic service composition systems.
Section 3 is a description of our ‘Affinity’ model for
dynamic service composition. This is the model that
needs to be validated. In Section 4, we describe the
Ambitious Traveler game which will serve as a means
to validate our model and which can be modified to
validate other dynamic systems as well. Section 5
discusses possible improvements and enhancements
which we envisage in the game in the future. Finally,
Section 6 concludes the paper.
2 RELATED WORK
In this section we discuss a few strategies for valida-
tion of dynamic systems that have been adopted by
researchers in the past.
Colombo et al. (Colombo et al., 2006) tackle the
issue of dynamic composition where the service el-
ements do not always behave along expected lines.
They provide an extension to the BPEL language in
the form of the ‘SCENE platform’ which addresses
this issue. The proposed platform is validated form-
ing an application using a set of real services and ob-
serving the behaviour of the application.
Silva et al. (Silva et al., 2009) propose the Dy-
namiCos framework which incorporates the require-
ments of different customers to dynamically put to-
gether personalized services. To validate the proposed
framework they put together an extensive prototype of
the framework which enables services to be deployed
and be published in a UDDI-like registry.
In their paper Eid et al. (Eid et al., 2008) de-
scribe a set of benchmarks against which to assess
various frameworks of dynamic composition. The set
of benchmarks is comprehensive and is broadly cat-
egorized into three parts: input subsystem, compo-
sition subsystem, and execution subsystem. A com-
position model or framework that scores well against
these benchmarks is considered good.
Shen et al. (Shen et al., 2007) present the Web
Service, Role and Coordinator (WSRC) model to han-
dle dynamism in web service compositions. In this
model, the process of service composition is divided
into three layers: Service, Role, and Coordinator. To
validate the model, the authors describe a case-study
of a vehicle navigation system which comprises a
global positioning system and a traffic control service.
These are a few of the validation techniques in use
for dynamic composition models and frameworks. Of
these techniques, the ones that are accurate are quite
complex like the prototype that the authors of Dynam-
iCos put together or the service composition put to-
gether out of real services for the SCENE platform.
On the other hand, the validation techniques that are
simple like the case-study in the WSRC model seem
ineffective.
The validation technique proposed in this paper,
the Ambitious Traveler game, on the other hand is
quite simple as it comprises an elementary game with
straightforward steps and instructions. Despite this, it
is quite precise and is able to quantitatively express
the performance of the dynamic composition model.
3 THE ‘AFFINITY’ MODEL FOR
SERVICE COMPOSITION
In this section we describe our ‘Affinity’ model for
service composition. It is specifically for the vali-
dation of this composition model that the Ambitious
Traveler game (described in Section 4) is used. The
game can, however, be modified appropriately and be
used for the validation of other dynamic systems as
well.
3.1 Assumptions on the Service Domain
In our ‘Affinity’ model for service composition, cer-
tain simplifying assumptions have been made about
the service domain from which the service elements
are selected. First, our approach assumes that the ser-
vice elements taking part in the composition process
are already available in the service domain. The ap-
proach therefore does not look into the issue of ‘dis-
covering’ the service elements. The reader interested
in the process of service discovery is pointed to (Wu,
2008) for an exhaustive discussion on the subject.
The potential service elements, which are as-
sumed to have been already discovered, are repre-
sented as arranged in a tiered fashion as shown in
Figure 1. Each tier or level corresponds to a specific
functionality of the composite application. All the
discovered service elements that correspond to the re-
spective functionality are placed at that level. The ar-
THE VALIDATION OF A DYNAMIC SERVICE COMPOSITION MODEL USING A SIMPLE GAME
261