Adaptation of Bees Algorithm for QOS-based Selection

and Replacement of Web Services

Rim Teyeb Jaouachi

, Maher Ben Jemaa

and Achraf Karray

ReDCAD Research Unit, National School of Engineers of Sfax, University of Sfax, Sokra Street, Sfax, Tunisia

Department of Computer Science and Applied Mathematics, University of Sfax, Sokra Street, Sfax, Tunisia

Keywords:

Web Services, Quality of Service, Replacement, Bees Algorithm, Optimization.

Abstract:

Web service is a new paradigm of internet software and distributed computing. With the growth of web

services number, the capability to ﬁnd the optimal service which can substitute the fault service from different

communities represents a very hard operation. The researched service must have the same functionalities of

the fault service and the highest quality of service scale. In this paper, we propose heuristic method based on

Bees Algorithm to ﬁnd the optimal service which can substitute the fault service. In our approach, we design a

distributed environment based on Peer-to-Peer architecture in order to present distributed communities which

are considered as ﬂowers and research requests as bees.

1 INTRODUCTION

Web services are used in various domains such as

business to business integration and web-based sys-

tems thats why the number of service webs consumer

is increasing. Therefore, the fault tolerance of web

services has become an active research domain in or-

der to make the use of this paradigm safe and improve

the performance and reability.

In the same context, we propose a method for pro-

viding fault tolerance in services oriented architec-

ture. Our approach focuses on building a distributed

architecture in which fails are detected automatically

and a suitable web service for substitution is found.

The selection of optimal service for substitution

represents a very hard operation because the large

space of solutions and multiple criteria of quality of

service are considered at once such as performance,

security, etc. Therefore, we propose the application of

heuristic method called Bees Algorithm to facilitate

research procedure of optimal service from a huge

number of web services grouped in communities.

This new technique is inspired from the way bees

forage the food and has been successfully applied to

various optimization domains and problems. This pa-

per is organized as fellow: the second section presents

web services; the third section shows the details of

our approach; in the ﬁrth section, we present the im-

plementation and the conclusion is a subject of last

section.

2 WEB SERVICES

2.1 Mechanisms of Discovered Web

Services

The discovery of the Web services represents the pro-

cess allowing the localization of the documents de-

scribing a service Web (WSDL ﬁle). However, with

the permanent increase in the number of the Web ser-

vices, the task of discovery becomes a difﬁcult task.

To facilitate the discovery of the Web services,

several approaches were deﬁned. These various ap-

proaches can be classiﬁed in two main categories:

discovered based on sequential algorithms and oth-

ers based on the heuristic ones. Among approaches

which tack part of the ﬁrst type we found WSDA

(Hoschek, 2002), Speed-R (Vu et al., 2005), PSWSD

(Vu et al., 2005), LARKS(Sycara et al., 2002), OWL-

S(Srinivasan et al., 2004), AASUD (Palathingal and

Chandra, 2004), etc.). For the second type of algo-

rithm, we found the approach (Jaeger and Mhl, 2007)

which uses the genetic algorithm to select a Web ser-

vice element of a composition while taking into ac-

count the maximization of the service quality of the

obtained composition.

2.2 Community of Web Services

There are several services which have the same func-

459

Teyeb Jaouachi R., Ben Jemaa M. and Karray A..

Adaptation of Bees Algorithm for QOS-based Selection and Replacement of Web Services.

DOI: 10.5220/0004083804590462

In Proceedings of the 7th International Conference on Software Paradigm Trends (ICSOFT-2012), pages 459-462

ISBN: 978-989-8565-19-8

 2012 SCITEPRESS (Science and Technology Publications, Lda.)

tionalities. However, these services are distinguished

from/to each other by their non-functional proper-

ties such as the performance, the cost and the safety.

Thus, in order to facilitate the research of the adapted

Web service, the services are classiﬁed by categories

or ﬁelds from where approach containing community

(Ran, 2003).

The principal goal of the community of services

is to facilitate the research and the selection of the

Web services. A community is characterized by its

abstracted interface, its model of quality and the vari-

ous recorded Web.

Contrary to a directory UDDI, a community is

characterized by a speciﬁc ﬁeld i.e. it gathers a whole

of services which are differed between them by the

non-functional properties. Thanks to a model of qual-

ity speciﬁc to each community, these parameters are

mentioned. A community is exposed through an ab-

stract interface which makes it possible to deﬁne the

interactions being able to begin with the community

services.

3 THE PROPOSED METHOD

Our purpose is to propose a new approach which opti-

mizes the time needed to replace the fault web service.

With growing and competing demands and invo-

cations, the probability of fault increase and the use

of fault tolerance mechanism become indispensable.

So that, we build a new architecture based on com-

munities of web services. To ﬁnd the optimal service

which can substitute the fault one, we use the Bees

Algorithm as an intelligent way of research.

3.1 Architecture

Our architecture consists of:

• A Layer for Semantic Clustering Web Services in

Community. Each community has a description

ﬁle which indicates its business.

• A Supervision Module. This component controls

the quality of service offered by different web ser-

vices in execution. If this level is below a cer-

tain value, it will invocate a module to choose an

equivalent service for substitution.

• A Substitution Module. This module is triggered

after a fault detection of a web service. It ensures

the selection of a web service offering the same

functionalities offered by the failed one and hav-

ing the highest level of quality of service.

So it is used, in the ﬁrst place, to ﬁnd the commu-

nity which has the same category with the failed

web service, and to choose between different ser-

vices web present in the community. The selec-

tion is based on quality of service.

• A Cache of Equivalence. This cache saves the

web services and their Web services equivalent.

It can make the search procedure faster than that

performed by the substitution module if both (de-

fault web service, web service equivalent) exists

in the cache. Each couple in this table has a pe-

riod of validity. After the expiration of this period

the associate couple will be removed.

3.2 Adaptation of Bees Algorithm in

Substitution Module

When searching for an equivalent web service, the

substitution module should choose from a wide range

of services already developed. To reduce the research

area, and to ﬁnd the desired web service, we will use a

new optimization technique called ”Bees Algorithm”.

3.2.1 The Bees Algorithm

As mentioned above, the Bees Algorithm represents

a new optimization algorithm proposed by Pham et al

(Pham et al., 2006). It is based on natural foraging

behavior of honey bees.

Due to its successful results in various domains

such as clustering problems, artiﬁcial neural network

training. This algorithm is considered as one of the

most important optimization algorithms.

3.2.2 Bees in Nature

The foraging process begins in a colony by sending

scout bees to search for ﬂower patches. They move

randomly from one ﬂower to another in order to ex-

ploit a large number of food sources. When they re-

turn to the hive, those scout bees deposed their nectar

then they evaluate the various plots visited and sort

them above certain quality threshold which can be

measured as a combination of some constituents such

as sugar.

They choose the (m) best patches and (e) top-rated

patches.

After this step, Scout bees perform a dance called

waggle dance.

This dance is essential for communication in the

colony. It provides three types of information: direc-

tion of a ﬂower patch, distance from the hive and a

quality of nectar.

This information helps the colony to send its

bees to collect nectar without the use of guide or

ICSOFT2012-7thInternationalConferenceonSoftwareParadigmTrends

460

maps. More follower bees are sent to more promis-

ing patches in order to gather nectar efﬁciently and

quickly.

3.2.3 The Basic Bees Algorithm

Listing 1 shows the pseudo-code of this algorithm.

More details are described in (Pham et al., 2006). To

execute this algorithm, a number of parameters must

be initialize namely: number of scoot bees (n), num-

ber of sites selected out of n visited sites (m), number

of best sites out of n visited sites, number of patches

selected for neighborhood research (m), number of

bees recruited for (m-e) selected sites (nsp) and stop-

ping criterion.

The algorithm begins with placement of (n) scoot

bees, randomly, in the search space.

3.2.4 Proposed Method for Substitution Module

The proposed method adapts the efﬁcacy of the Bees

Algorithm to optimize the research procedure. When

searching for an equivalent web service having the

same functionalities of failed web service, the sub-

stitution module should choose from a wide of web

services already developed.

To minimize research area and to ﬁnd a desired

web service in the shortest period, we use this tech-

nique. We consider a bee as a web service query and

a community of web services as a patch.

Our proposed approach based on Bees Algorithm

consists of ﬁve steps.

• Step1: random exploration. This step consists to

send search queries in parallel way to different

communities of web services. The choice of com-

munities at this stage is randomly.

• Step2: measure of similarity. In step 2, the algo-

rithm evaluates the degree of similarity between

business domain of failed service and visited com-

munities. Each community has a value of ﬁtness

function which indicates the degree of semantic

similarity.

• Step3 and step4: Sorting and selection of suitable

community. In step 3, visited communities are

sorted. according to similaritys degree collected

from the second step.

• In step 4: A community having the highest degree

of similarity is selected as elite community.

• Step5: Exploration of elite communitys neighbor-

hood To explore the neighborhood of elite com-

munity, the algorithm sends research queries to

community selected in step 4.

Obtaining maximal similarity between two areas

of interests: area of fault web service and the selected

community represent the stopping criterion.

4 IMPLEMENTATION

In order to implement our approach different tools are

used.

4.1 JXTA Platform

JXTA (Vogel et al., 1995) is a project developed by

Sun for the Peer-to-Peer (P2P). It used to facilitate

the exchanges between peers on a network.

This platform offers several services such as:

• Dynamic search.

• Resource sharing (documents, ﬁles...).

• Creation of peers groups.

• Communication in a protected way.

• Collaboration with other applications.

• Etc.

Based on standards such as TCP/IP, HTTP and

XML, JXTA is independent of any programming lan-

guage, of any protocol of transport and any operat-

ing system. The choice of this platform is adequate

for implementing our distributed communities net-

work. Each community can be presented by a group

of JXTA peer. To present a distributed community

environment, we implement several peer groups, i.e.,

registry of education, sport, and travel.

4.2 WordNet

WordNet represents a semantic network of the lexical

data bases. Its goal is to index, classify and connect

the semantic contents and lexical of the English lan-

guage (complete version).

WordNet was also developed for other languages

like French (represents only 25

• Synsets: when the direction of a word is deﬁned

by a whole of the synonyms and a deﬁnition

• Lexical Relations: when the direction of a word is

deﬁned by a whole of the semantic relationships

to other directions (Example chien/caniche, etc)

4.3 WordNet: Similarity

WordNet: Similarity is free software which measures

the similarity between two terms. It uses three mea-

surements of semantic similarity and six measure-

ments of linguistic proximity

AdaptationofBeesAlgorithmforQOS-basedSelectionandReplacementofWebServices

461

• Measurements of Linguistic Proximity of Word-

Net: is based on the analysis of the hierarchy

WordNet ” is a ” Several methods using this prin-

ciple of measurement such as measurements of

WuP (De la Rosa, 2004).

• Measurements of Semantic Similarity of WordNet:

is not based only on the relation ” is a ” of Word-

Net, but they use other non hierarchical relations

• Three Measurements of Semantic Similarity Of-

fered by WordNet: is: Vector (Leacock and

Chodorow, 1998) Hso (Patwardhan, 2003) Lesk

(Hirst and St-Onge, 1997).

These two tools are used in order to Measure the

similarity between business domains of failed web

service and visited communities.

5 CONCLUSIONS

In this paper, we adapted a new technique of opti-

mization called Bees Algorithm. This algorithm opti-

mizes web services discover and the selection of suit-

able web service which can substitute fault one.

In our approach, we implemented a peer to peer

environment in order to present distributed commu-

nities. Our method is based on two important steps.

The ﬁrst step is to ﬁnd a community having the

same/equivalent business domain of fault one with

minimization of research procedures time. The sec-

ond step is to calculate quality of services level for all

web services present in founded community and to se-

lect the best between them offered the highest score.

REFERENCES

De la Rosa, R. (2004). Dcouverte et Slection de Services

Web pour une application Mlusine. Master’s thesis,

Universit Joseph Fourier, Grenoble.

Hirst, G. and St-Onge, D. (1997). Lexical chains as repre-

sentations of context for the detection and correction

of malapropisms.

Hoschek, W. (2002). The web service discovery architec-

ture. In Proceedings of the 2002 ACM/IEEE confer-

ence on Supercomputing, Supercomputing ’02, pages

1–15, Los Alamitos, CA, USA. IEEE Computer Soci-

ety Press.

Jaeger, M. C. and Mhl, G. (2007). Qos-based selection of

services: The implementation of a genetic algorithm.

In In KiVS 2007 Workshop: Service-Oriented Archi-

tectures und ServiceOriented Computing (SOA/SOC),

pages 359–370.

Leacock, C. and Chodorow, M. (1998). Combining local

context and WordNet similarity for word sense identi-

ﬁcation, pages 305–332. In C. Fellbaum (Ed.), MIT

Press.

Palathingal, P. and Chandra, S. (2004). Agent approach for

service discovery and utilization. In Proceedings of

the 37th Annual Hawaii International Conference on

System Sciences (HICSS’04).

Patwardhan, S. (2003). Incorporating Dictionary and Cor-

pus Information into a Context Vector Measure of Se-

mantic Relatedness. Master’s thesis, University of

Minnesota, Duluth.

Pham, D. T., Ghanbarzadeh, A., Koc, E., Otri, S., and Zaidi,

M. (2006). The bees algorithm a novel tool for com-

plex optimisation problems. In Proceedings 2nd Vir-

tual International Conference on Intelligent Produc-

tion Machines and Systems IPROMS, pages 454–459.

Ran, S. (2003). A framework for discovering web services

with desired quality of services attributes. In ICWS,

pages 208–213.

Srinivasan, N., Paolucci, M., and Sycara, K. P. (2004). An

efﬁcient algorithm for owl-s based semantic search in

uddi. In SWSWPC, pages 96–110.

Sycara, K., S, W., M, K., and Lu, J. (2002). Larks: Dynamic

matchmaking among heterogeneous software agents

in cyberspace. In in Cyberspace. Autonomous Agents

and Multi-Agent Systems, pages 173–203.

Vogel, A., Kerherv, B., von Bochmann, G., and Gecsei, J.

(1995). Distributed multimedia and qos: A survey.

IEEE Multimedia, 2:10–19.

Vu, L.-H., Hauswirth, M., and Aberer, K. (2005). To-

wards p2p-based semantic web service discovery with

qos support. In Business Process Management Work-

shops, pages 18–31.

ICSOFT2012-7thInternationalConferenceonSoftwareParadigmTrends

462