A Web Service Discovery Approach Based on Hybrid Negotiation

Raja Bellakhal

, Walid Chainbi

and Khaled Ghédira

Higher Institute of Business, University of Tunis, Tunis, Tunisia

National Engineering School of Sousse, University of Sousse, Sousse, Tunisia

Keywords: Web Service Discovery, Negotiation, Quality of Service (QoS), Multi-Agent-Systems (MAS).

Abstract: An effective discovery system must be able to retrieve services responding to the users’ specific preferences

in a changing and dynamic environment. Actually, the existing discovery systems have three problems.

Firstly, some of them fail to find Web services providing the same QoS as defined in their related

description files, since the QoS data are considered as static. Secondly, the discovery systems based on

negotiation lack the accuracy in simulating similarly the real humans’ interactions. Thirdly, the negotiation

based approaches implemented to discover services are static and do not consider contexts as well as

characteristics of each provider. These shortcomings affect negatively the systems performance and

usability. Consequently, the quality of the returned services as well as the systems’ reputation will be

deteriorated. In this paper, we propose an hybrid discovery approach based on negotiation that solve these

drawbacks. We argue that our approach enhances the discovery system performance and usability by

implementing a negotiation process that is closer to humans’ interactions. Moreover, by considering the

existing dependencies between the concurrent negotiations, the discovery process will be more efficient.

Unlike previous work, the negotiation process is dynamic by taking into account the provider’s context and

reputation.

1 INTRODUCTION

The Web is no longer a simple way to access

information, its evolution makes it also a provider of

services. Web Services refers to applications

exposed by providers over the Web. They provide a

standard means of interoperating between different

software applications, running on a variety of

platforms and/or frameworks. Service discovery is

the process of locating Web service, and retrieving

Web services descriptions that have been previously

published (Bromuri et al., 2009). Often Web

services operate in a rapidly changing environment.

Consequently, their QoS parameters change quite

frequently, either due to the external (e.g., the

variance on the service demand, server outages, the

congestion at the router (Bannock Consulting,

2000)) or the internal (e.g., poor upstream

connectivity (Kouki et al., 2013), the changes of the

providers’ preferences (Bannock Consulting, 2000))

environment changes. In most studies dealing with

the discovery task, QoS parameters are considered

as static (Ran, 2003), (Zhou and Chen, 2005),

(Lakhal and Chainbi, 2012). Therefore, the non-

functional features of the returned services will not

be the actual ones that reflect the service

functioning. Some studies have tried to find

solutions for the static QoS parameters problem

(Chen et al., 2010), (Zhou and Chen, 2009), (Platzer

et al., 2007). These solutions use respectively

statistics (e.g., the sampling method (Zhou and

Chen, 2009)) and networks (e.g., sniffing, probing,

and interceptors (Platzer et al., 2007)) based

methods. These technologies need over resources

utilization as well as huge computational and

statistical efforts. Recently, other works adopting the

negotiation approach in order to dynamize the QoS

parameters appeared (Bentahar et al., 2008), (Napoli

et al., 2013). The adoption of negotiation while

discovering the best service has also several benefits

including the discovery time optimization, the

resolution of conflicts between the providers’ and

clients’ preferences, and the improvement of the

success rate. However, the main weakness of these

solutions is related to the difference between

humans’ behaviors during their interactions and

those of agents during the systems functioning.

These systems do not simulate some common

315

Bellakhal R., Chainbi W. and Ghédira K..

A Web Service Discovery Approach Based on Hybrid Negotiation.

DOI: 10.5220/0004940803150322

In Proceedings of the 4th International Conference on Cloud Computing and Services Science (CLOSER-2014), pages 315-322

ISBN: 978-989-758-019-2

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

negotiators’ behaviors observed during their real life

interactions. Simulating these details about

negotiators’ behaviors will positively impact the

system’s performance and outputs. During real life

negotiation, each client uses his/her experience,

competences, and knowledge to negotiate differently

his/her goals. Also, negotiators make continuously

new offers and give explanations about the reasons

of choosing such offers. In our approach, negotiators

will generate offers, give explanations and behave

differently. Additionally, in these systems the

negotiation strategies are static. These strategies are

used by all providers without caring about the

specific context and characteristics of each one of

them. Also the consideration of dependencies

between the different negotiation processes is

missed. These dependencies are the cooperation

relations existing between the different client’s

instances. These instances exchange information

about the providers’ offers and use them to propose

more attractive offers. Considering these

dependencies make more efficient the overall

negotiation process. The objective of this paper is to

present a discovery approach based on an hybrid and

a dynamic negotiation process. Given that they are

software components that have autonomous

behavior and that try continuously to meet their own

objectives, by taking into account their resources,

skills and perceptions, agents are suitable entities to

be adopted for the simulation of both users’ and the

providers’ behaviors during their interactions

(Chaib-draa et al., 2001).

The reminder of this paper is organized as

follows. In the second section, we review the

existing discovery approaches, and we outline their

main characteristics as well as their drawbacks. In

the third section, we highlight the main contributions

brought by our approach. A buying-selling case

study that illustrates our approach is presented in the

last section. Finally, we summarize our work and we

give hints concerning our future work.

2 RELATED WORK

Two types of discovery approaches have been

proposed to solve the static QoS parameters

problem, namely, the discovery approaches that are

network as well as statistical based methods, and

those that are adopting the negotiation. Compared to

researches dealing with the first type of approaches,

the negotiation based works are fewer.

2.1 Approaches based on Network and

Statistical Methods

In (Al-Masri and Mahmoud, 2009) a WSB (Web

Service Broker) is used to update the QoS

parameters values. More especially, a WSCE (Web

Service Crawler Engine) is used to collect

dynamically the updated QoS data. In (Julie and

Kumar, 2012), a collaborative filtering approach that

predicts automatically the QoS parameters values is

adopted. An important drawback related to this work

is that there are no large-scale real-world Web

service QoS data sets available for studying the

prediction accuracy. In (Zhou and Chen, 2009) the

data related to the users’ and the providers’

feedbacks are gathered and then processed according

to a set of statistical methods. The drawback of

approaches which consider the users’ and the

providers’ feedbacks is related to the big statistical

efforts computations. Other works use statistical

methods to estimate values of the QoS parameters

before the reception of feedbacks (Thio and

Karunasekera, 2005). In these approaches, the most

statistical jobs are executed when original data is

received in the service registry. Consequently, this

will take up many other hardware resources of the

service registry (Zhou and Chen, 2009). In (Chen et

al., 2010), the authors present a model for the

management and the discovery of services based on

two dynamic QoS parameters, namely the response

time and the execution time. The monitor that acts as

an interceptor during each service invocation is

used. Most of the existing works enabling the

dynamicity of the QoS data that are discussed above

are suffering from general drawbacks. Many of these

approaches are restricted in the sense that they can’t

handle the dynamicity of some QoS parameters

types including the service’s cost and the service’s

price. These parameters are of a paramount interest

in the cloud computing field, and should be

considered as dynamic during the discovery process.

Additionally, some of the presented approaches

using statistical methods require over resources

utilization as well as huge computational efforts.

Moreover, since some systems use the gathered

information during the service invocation to keep

up-to-date, then in the case where a service is not

invoked periodically, its related non-functional

parameters can’t reflect the actual service’s

performance.

2.2 Approaches Based on Negotiation

Bentahar et al., adopt in the context of the

CLOSER2014-4thInternationalConferenceonCloudComputingandServicesScience

316

communities’ construction, an argumentation based

approach (Bentahar et al., 2008). The selection of

the best service is done in terms of functionality and

QoS (i.e., response time, availability, throughput,

reliability, etc.). Agents use knowledge, beliefs, and

argumentation capabilities to reason. A persuasive

negotiation protocol composed of different dialogue

game was used. A dialogue game is the set of moves

that are made during the negotiation. This work

adopts only one model of negotiation and the

complexity of the used protocol depends on the size

of the agents’ knowledge bases. A negotiation

approach that adopts a coordinator between the

different Web services based on the MAIS

(Multichannel Adaptive Information Systems)

architecture is presented in (Comuzzi and Pernici,

2005). The negotiation capabilities related to each

provider are presented as policies in the service

specification. Services involved in the negotiation

are selected, and then one auction based strategy is

chosen from the different auction models and

applied between these services. In this work, only

one negotiation model is used and the implemented

system is not based on the agent technology. The

work presented by Napoli et al., is close to our work,

since it deals with the static QoS parameters

problem (Napoli et al., 2013). However, unlike our

work, the generation of offer related to the QoS

parameters is permitted only in the provider’s side,

the user has only the role of evaluating offers. In this

work, the standard contract net protocol and a

unique negotiation model are adopted. The paper

presented by Siala and Ghédira use a variant of the

contract-net protocol, called the directed contract-net

protocol that has the effect of decreasing the number

of the negotiation rounds comparing to the standard

Contract-Net while composing services. A

coordinator supervising the different negotiations is

used. Dependencies between the different

negotiations strategies are not considered in this

work (Siala and Ghédira, 2011). Bromuri et al.,

present a decentralized multi-agent system that

adopts an argumentation based strategy to find the

best service. In this work, the best service is firstly

selected among a set of the available services, then a

negotiation session is started with the chosen service

in order to achieve an agreement about the required

QoS parameters. This strategy restricts the client’s

chances to find the best offer since he/she negotiates

his/her preference with only one provider, while

other providers can have more interesting offers

(Bromuri et al., 2009).

3 THE HYBRID NEGOTIATION

APPROACH CONTRIBUTIONS

Given the promising solutions brought by

negotiation, recent discovery approaches based on

negotiation models arise. A set of these works have

been already presented in the related work section.

These works have three common problems. Firstly,

the negotiation systems presented by these works

lack the realism in simulating perfectly the real

human interactions and behaviors during the

negotiation process. In fact, only a unique

negotiation strategy based on one negotiation model

is adopted for simulating providers’ and clients’

behaviors. In real life each client and provider has

his/her own strategy to negotiate. In these works, all

negotiators are obliged to follow the same strategy

in order to reach their different goals. In this paper,

our aim is to relax this assumption by introducing a

system which applies more than one strategy based

on different negotiation models. In our work, we

adopt the argumentative and the heuristic models.

The formal description of the adopted strategies is

out of the scope of this paper, since our main interest

is to present our discovery process based on hybrid

negotiation. Secondly, all these works assume that

the negotiation process is static i.e., once the

negotiation strategy is defined by developers it will

be applied similarly with all providers without

considering their different context and/or

characteristics. Finally, dependencies between the

different negotiations are missed. Each bilateral

negotiation involved in the whole negotiation

process is conducted regardless to other processes

which result in longer negotiation process and sub-

optimal offers. In our study, we solve these

problems by considering three main contributions.

The first contribution is to make more realistic the

negotiation process by simulating some details

observed during the real humans’ interactions. The

second contribution is to make dynamic the

negotiation process. The final contribution is to

consider dependencies between the concurrent

bilateral negotiations.

3.1 The Simulation of the Real

Humans’ Interactions

The simulation of humans’ interactions implies the

creation of an artificial world of agents that are in

interaction and that have humans’ capabilities to

interact with their environment and to react

according to their own perception, goals, and beliefs.

AWebServiceDiscoveryApproachBasedonHybridNegotiation

317

Since systems are substituting the real human

interactions, they have to be as much as possible

close to the humans’ behaviors as well as to their

strategies during the negotiation. The second feature

observed in human interactions, but that is missed in

the current systems dealing with the discovery of the

best service is the exchange of explanations between

the different parties. In our work, we solve conflicts

among negotiators by enabling the two parties to

exchange explanations through arguments. In works

dealing with the service discovery based on

argumentative based models the explanatory role is

not yet studied. The third feature characterizing the

real human interactions that is missed in some of the

existing works is related to the providers’ and

clients’ role in generating offers. Information about

the opponent’s offers is used to propose more

attractive offers during the next negotiation rounds.

3.1.1 An Hybrid Strategy

During their daily real business interactions, sellers

and buyers are using different strategies to convince

each others about their offers and preferences. Each

negotiator has his/her own way to convince his/her

opponent and to generate offers. Obliging all

providers and all clients to adopt the same strategy is

not possible since some of them haven’t essential

information and backgrounds to use systems based

on some kind of negotiation models. For example, if

the negotiation system implements an experience

based strategy, users who haven’t yet experience and

who want to purchase the service for the first time,

will find problems in using this system. Also other

models such as the argumentation based model

require a predefined set of knowledge and

arguments. However, sometimes users aren’t able to

formulate readily suitable arguments to convince

their opponents and to express fluently all their

preferences. Implementing different strategies will

improve the discovery system usability level.

Usability is defined by three features namely, the

system’s effectiveness, the system’s efficiency, and

the user’s satisfaction (Frøkjær et al., 2000).

Effectiveness is related to how much the system is

able to respond the users’ requests. Satisfaction is

related to the comfort and subjective user’s

evaluation of the interaction. The efficiency is

related to how the system is able to achieve the best

result with less effort and the minimum time. A

system qualified by a high usability level is better

suited to be consumed by different types of users.

Along with such a system, users having different

requirements are more luckily to find the

functionalities and the conditions helping them to

discover their services. Consequently, the reputation

of the discovery system will be improved. When

adopting different strategies we try to balance

between the cost and the effectiveness. The

argumentative model is powerful enough to achieve

an agreement but it’s costly in terms of consumed

resources and complexity. While a less powerful

model such as the heuristic model can improve the

complexity and decrease the chances to achieve an

agreement. Moreover, contexts and attitudes of each

provider are different, so it’s difficult to apply the

same strategy with all providers. The client has to

collect information about the provider’s attitude and

accordingly he/she must choose the adequate

strategy that complies with the providers’

characteristics.

3.1.2 The Role of Clients and Providers in

Giving Explanations

The provider’s role in bringing explanation for users

about some aspects that may be not obvious for

them is missed within the existing discovery

approaches. This is possible with the negotiation via

the adoption of the explanatory arguments. For

example, the client can ask the provider if it’s

possible to have a discount or a higher service

availability level. Also he/she can inquire further

details about the service functionalities and about

the reason of uttering a given offer. Within the

negotiation mechanism, the provider is able to

respond to such requests and to give additional

explanations. In works adopting the argument based

models in order to select the best service, other

types of arguments are used, including rewards and

threats. These arguments have the effect of inducing

and convincing negotiators rather than giving

explanations. In the case of an explanatory argument

about the last offer, the opponent examines firstly

the argument and accordingly he/she can decide if

he/she will accept or refuse it. If the opponent’s

beliefs and preferences comply with the explanatory

argument then he/she will accept it. Otherwise, this

explanation will be used by the opponent during the

building of his/her next offers. The opponent will

have an idea about the proponent’s preferences and

beliefs, and accordingly he/she can formulate more

attractive offers to him/her. Explanatory arguments

have the role of resolving conflicts among the

negotiators’ preferences. Clients and providers use

them to explain the reason of refusing the

opponent’s offer. This information will be then

retained by the opponent and used to exchange more

CLOSER2014-4thInternationalConferenceonCloudComputingandServicesScience

318

attractive offers. This will make faster the discovery

process and helps to resolve gradually conflicts.

Also, the quality of the last offer will be improved,

since it will be the result of exchanging more

attractive proposals.

3.1.3 The Role of Clients and Providers in

Generating Offers

In some existing works, the provider has the role of

generating offers, while the user has only to evaluate

offers (Napoli et al., 2013). In our work, users and

providers generate and evaluate offers. Our main

interest is satisfying the users’ goals. To do so, users

must have their own strategies and tactics to

generate offers reflecting their preferences as is done

during the real human interactions. These proposals

will be used by providers as basis to the construction

of their next offers. Both parties must evaluate and

generate offers because this will make smarter and

faster the negotiation process. Evaluating only offers

doesn’t show the preferences of the negotiators,

rather it reveals only their positions. Thus, a rational

client respectively (provider) who wants to gain the

negotiation and to reach as faster as possible an

agreement must generate offers rather than

evaluating only the provider’s (the client’s) offers.

This will help both negotiators to construct faster an

idea about each other’s preferences, and then to

reach more rapidly an agreement.

3.2 The Dynamic Negotiation Process

Most of works dealing with the selection of the best

service by adopting a negotiation approach assume

that the negotiation process is static. In our work, we

make dynamic the negotiation process by selecting

the appropriate negotiation strategy to be adopted

with each provider. In fact, the choice of the best

strategy depends on different parameters including

the provider’s context, reputation, and preferences.

Before starting the negotiation, the system must

gather information about providers. Accordingly, the

best strategy to be adopted with these providers is

selected. For example, systems must have an idea

about the provider’s context such as, if the provider

is selling during a high or a sale season. In the latter

case, the provider is willing to make more

concession. Accordingly, the system can choose a

negotiation strategy that doesn’t need big efforts in

convincing the provider such as a heuristic based

strategy. However, if the provider is in a high season

which means that there is a big competition among

providers, then the system may choose a stronger

strategy such as an argument based strategy.

Reputation is also important. Reputation is

determined based on the attitude of the provider

during his/her past interactions. The system should

know if the provider is a stubborn negotiator or a

collaborative one. Accordingly the appropriate

negotiation strategy is adopted. Moreover, knowing

the provider preferences toward the values of the

QoS parameters can give us an idea about the

distance between the user’s offers and those of the

provider. This distance represents evidence about

how will be difficult the conversation with one

provider. If this distance is large, the client will find

difficulties to convince the provider. It will be easier

to convince each other if they have closer

preferences. In the first case, convincing the

provider will be a hard task; a negotiation strategy

based on strong arguments must be adopted.

However, in the second case, an experience or a

heuristic based model can sufficiently make the task.

By considering one of the latter information about

each provider, the best strategy that must be applied

with each provider is determined. Choosing the best

strategy that fits the provider’s profile, avoids for

example the adoption of less powerful negotiation

model with the more stubborn providers as well as a

heavy strategies with the less stubborn ones. In the

first case, we are increasing the chance of achieving

an agreement. However, in the second case we are

reducing the consumption of unnecessary resources.

3.3 The Dependency among the

Bilateral Negotiations

In the existing works, dependencies between the

different negotiations involved in the whole

negotiation process are not considered. In our work,

the relation between the concurrent bilateral

negotiation processes is ensured through an

intermediate party. During each negotiation round,

the intermediate party gathers information about

offers proposed by each provider from each client’s

delegate and communicates this information to other

client’s agent instances. In our work, we assume that

there is a competition between providers and also

cooperation relation between clients’ agent

delegates. Each provider’s aim is to present a more

attractive offer to the user. Each client’s agent is

able to be informed about other providers’ offers

through their correspondent clients’ agent instances.

The client’s agent will use this information to induce

the provider’s agent to make better offers than those

of its agent’s competitors. For example, if one of the

client’s agents knows that the other providers’

AWebServiceDiscoveryApproachBasedonHybridNegotiation

319

agents offered more attractive proposals, it will then

refuse the offer proposed by the provider’s agent

with which it is currently dealing. Also, it will

induce the provider to make more attractive offers.

Consequently the competition between providers

will be more intense. This competition will be in the

favour of the user and will accelerate the negotiation

process progression. Each provider will try to make

more interesting offers in order to get the deal. Thus,

providers and clients will exchange more and more

attractive offers until achieving an agreement. This

will speed up the negotiation process and scale down

distances between the negotiators’ preferences.

When the cooperation among clients’ instances and

the competition among providers are adopted

together, clients are luckiest to get at the end the best

offer.

4 A CASE STUDY

In this section, we give an example of a buying-

selling transaction that illustrates how the

negotiation protocols presented by Moschoyiannis et

al., (2009), and Venugopal et al., (2008) are applied.

We make some modification in these protocols to

show benefits of considering dependencies between

the bilateral negotiations. In future papers, we will

demonstrate through different types of experiments

the other contributions.

In this example, we consider that the client’s

agent aims is to benefit from a service with a low

price and high levels of availability, and reliability.

However, providers’ agents’ ambition is to have the

opposite.

4.1 the Protocols Description

In order to illustrate our hybrid approach, we use the

standard alternating-offers protocol and also the

protocol presented by Moschoyiannis et al.. The

latter protocol is composed of a set of moves which

are respectively, m1:request(negotiation), m2:refuse

(negotiation), m3: accept(negotiation), m4: offer(x),

m5: accept(x), m6: refuse(x), m7: challenge(x), m8:

argue(x), m9: withdraw(x). It specifies also when a

particular move is made in the course of a

negotiation dialogue by defining the pre-condition

and the post-condition of each move. For instance,

the move challenge can be uttered if only there is

beforehand an offer move. The post-condition of the

challenge move is to utter an argue message or a

withdraw message. We extend this protocol to make

possible the proposal of more than one message at

once. For example, for the message m4 we can reply

by an m7 and m4 messages, instead of m7 only.

Also we enable the utterance of an offer move after

an argue move. Also, we make changes in order to

adapt the protocol to the negotiation over the QoS

parameters rather than it is presented in the context

of the resource allocation. In the following, we

denote by A1 the first client’s delegate agent, and by

P1 its correspondent provider’s agent.

The standard alternating-offers is a bilateral

negotiation protocol. Negotiators exchange only

proposals, counter-proposals, accept and reject

messages. We denote by A2 the second client’s

delegate agent, and by P2 its correspondent

provider’s agent. Since the presented negotiation

protocols cover more than one potential alternative

scenario, we will be restricted to the more principal

negotiation scenario. Furthermore, we only describe

informally the proposals, arguments, and messages.

In this example, we keep static the availability and

the reliability parameters values and we focus only

on the price.

4.2 The Main Scenario

Step1

Negotiation 1

m1:Request(negotiation)

Negotiation 2

Offer (50 $, 0.8, 0.9)

Step2

Negotiation 1

m3:Accept(negotiation)

Negotiation 2

Counter-offer (59$, 0.8, 0.9)

Step3

Negotiation 1

m4:Offer(50 $, 0.8, 09)

Negotiation 2

Refuse(59 $, 0.8, 0.9)

Step4

Negotiation 1

m7: challenge (50 $, 0.8, 0.9).

-Why do you proposed as price 50 $, as availability

0.8, and as reliability 0.9?

- What argument you can give to let me believing

about this offer?

m4:Offer (58$, 0.8, 0.9)

CLOSER2014-4thInternationalConferenceonCloudComputingandServicesScience

320

Negotiation 2

Offer (58 $, 0.8, 0.9)

Step5

Negotiation 1

m8:argue(I need it for a daily use. I would like to

have a high availability and reliability level with a

reasonable price).

Negotiation 2

Counter-offer(54$, 0.8, 0.9)

Step6

Negotiation 1

-IF the m8 complies with the P1’ goals then it will

accept the P1’ offer. And the process ends with an

agreement.

-ELSE, P1 refuses the A1’ offer. End of the process

without agreement. P1 can give a proposal.

m4: Offer (57 $, 0.8, 0.9)

Negotiation 2

Offer (56 $, 0.8,0.9 )

Step7

Negotiation 1

m7: challenge (57 $, 0.8, 0.9)

-Why do you proposed as price 57 $, as availability

0.8, and as reliability 0.9?

Negotiation 2

Accept(56 $, 0.8,0.9)

Step8

Negotiation 1

m8: argue ({If you accept this proposal, you will

have a discount}).

Step9

Negotiation 1

m4: Offer (55 $, 0.8, 0.9)

Step10

Negotiation 1

m5:Accept (55$, 0.8, 0.9).

In step1, the agent A1 makes a request move.

Simultaneously, the client’s agent A2 makes an

initial offer. In the second step, P1’s agent accepts

the A1’s request, and the negotiation starts. In the

second bilateral negotiation, agent P2 doesn’t agree

with the A2’s offer and makes a counter-offer. In

step 3, the A1’s agent generates an offer according

to its preferences and beliefs. In parallel, A2 refuses

the last offer of P2. In step 4, P1 makes a challenge

move, and it proposes a first offer. In parallel, P2

proposes a cheaper price than that it offered in the

previous step. In step 5, A1 sends to P1 an

explanatory argument in which it explains its reason

of making such an offer. Indeed, the client’s agent

wants to have a high availability and reliability

levels in addition to a low price since the client will

consume daily the service. This explanatory

argument helps provider’s agent to change its

anterior beliefs to be more flexible and closer to the

client’s preferences. In fact, closer the agents’

preferences, the least conflicts are. In the second

negotiation, the A2’s agent makes a concession and

proposes 54 $ as a price. In step 6, the intermediate

party’s agent gathers information about the last

providers’ offers and communicates them to the

other clients’ delegates. Each delegate, during each

round must inform its opponent’s agent about this

information and induce it to make more interesting

offers. It also tells its opponent about the offers that

were refused by the other client’s agents. According

to its policies and strategies, each provider’s agent

decides which offer to propose. For example, in the

step6, P2 proposes as a price 56$ which is lower

than 58$, proposed by P1 during the step4. It’s a

time lost for the provider’s agent to propose an offer

that is less pertinent to the client. By knowing the

offers refused by the other clients’ instances, the

provider’s agent can understand better the client’s

preferences. Consequently, it can propose more

interesting offers. For example in the step4, P1

avoids to offer as a price 59 $, since this proposal

was refused by the other client’s agent in the step3.

The consideration of the relation between the two

bilateral negotiations makes faster the achievement

of an agreement. In step 7, A1 makes a challenge

move. In parallel A2 accepts the P2’s offer, and the

second negotiation process ends. In step 8, P1 makes

a reward argument in which it promises A1 a

discount if it accepts the offer. According to its

goals, A1 can either accept or refuse this argument.

If it accepts the argument, then an agreement is

achieved, and the client will have a discount. In step

9, A1 doesn’t accept the P1’s argument, and it

proposes 55$ as a price. In step 10, P1 examines the

A1’s last offer and according to its preferences and

beliefs it decides if it will accept, refuse the offer or

make a withdraw move. In our case, P1 accepts the

A1’s offer, and an agreement is achieved. Between

the two final offers, the coordinator’s agent will

choose the one which decrease the price, which is in

this case the P1’s offer.

AWebServiceDiscoveryApproachBasedonHybridNegotiation

321

5 CONCLUSION

Changes on the Web service environment and on

users’ requirements have led to new works having as

goal the establishment of discovery systems that can

respond to users’ requests and assure the credibility

of results. In this context, several approaches have

been applied including the negotiation which leads

to significant results related mainly to the

improvement of the success rate and the

optimization of the search time. In this paper, we

have introduced new aspects which are missed in

these approaches. The first aspect is related to the

consideration of the real life negotiation

characteristics during the simulation. These

characteristics are the use of different negotiation

strategies during interactions, the exchange of

explanations between negotiators, and the role of

clients and providers in generating offers. The

adoption of an hybrid strategy makes the discovery

system more realistic and oriented to the largest

number of users. The second contribution is to make

dynamic the negotiation process. This has the effect

of making the negotiation more efficient by

choosing for each provider the most suitable

negotiation strategy that fits his/her reputation,

preferences and context. The third contribution is the

consideration of dependencies among the concurrent

negotiations. This will impact the quality of the final

offer which will be also the result of the competition

between providers. Moreover, this will make faster

the discovery process since the client will avoid the

evaluation of offers that weren’t accepted by other

clients’ instances during the past rounds. Also the

providers will avoid proposing offers that were

refused by the other clients’ instances during the

previous rounds. Our future work, will be devoted

initially to the presentation of a formal definition of

our negotiation approach and secondly for the

development and the experimentation of our system.

REFERENCES

Bannock Consulting, 2000. Quality of Service Parameters

for Internet Service Provision, European Commission,

DG Information Society.

Frøkjær E., Hertzum M., and Hornbæk K., 2000.

Measuring usability: are effectiveness, efficiency, and

satisfaction really correlated, In Proc. CHI,

Washington.

Chaib-draa B., Jarras I., and Moulin B., Chaib-draa B.,

2001. Systèmes multiagents: Principes généraux et

applications, In: Agent et Systèmes Multi-Agents,

Hermès, Paris.

Ran S. P., 2003. A Model for Web Services Discovery

with QoS, SIGecom Exchange.

Thio N., and Karunasekera S., 2005. Automatic

Measurement of a QoS Metric for Web Service

Recommendation, the 2005 ASEC.

Comuzzi M. and Pernici B., 2005. An Architecture for

Flexible Web Service QoS Negotiation, EDOC 2005.

Zhou C. and Chen H., 2005. An Objective and Automatic

Feedback Model for Qos Evaluation, ICIS '09

Proceedings of the 2nd ICIS, ACM New York.

Platzer C., Rosenberg F., and Dustdar S., 2007. Enhancing

Web Service Discovery and Monitoring with Quality

of Service Information, In: Securing Web Services:

Practical Usage of Standards and Specifications, Idea

Publishing Inc.

Bentahar J., Maamar Z., Wan W., Benslimane D., Thiran

P., and Subramanian S., 2008. Agent-based

communities of Web services: An argumentation-

driven approach, Service Oriented Comput. Appl., vol.

2, no. 4, pp. 219–238.

Venugopal S., Chu X., and Buyya R., 2008. A Negotiation

Mechanism for Advance Resource Reservation using

the Alternate Offers Protocol. Proceedings of the

(IWQoS 2008).

Zhou C., and Chen H., 2009. An Objective and Automatic

Feedback Model for Qos Evaluation, ICIS '09

Proceedings of the 2nd ICIS.

Al-Masri E., and Mahmoud Q. H., 2009. A Broker for

Universal Access to Web Services. 2009 Seventh

Annual Communications Networks and Services

Research Conference.

Bromuri S., Urovi V., Morge M., Stathis K., and Toni

F., 2009. A multi-agent system for service discovery,

selection and negotiation. AAMAS.

Albreshne A., Fuhrer P., and Pasquier J., 2009. Web

Services Technologies: State of the Art, University of

Fribourg, Switzerland.

Moschoyiannis S., Krause P., Bryant D. and McBurney P.,

2009. Verifiable Protocol Design for Agent

Argumentation Dialogues In: 3rd IEEE ICDET,

Istanbul, TURKEY.

Chen H-P., Zhang C., and. Yang G., 2010. A Model for

Managing and Discovering ServicesBased on

Dynamic Quality of Services. Journal of Networks.

Siala F., and Ghédira K., 2011. A multi-agent selection of

web service providers driven by composite qos. In

ISCC.

Julie D., and Kumar K., 2012. Optimal Web Service

Selection Scheme With Dynamic QoS Property

Assignment. IJART, Vol. 2 Issue 2.

Ben Lakhal R., and Chainbi W., 2012. A Multi-Criteria

Approach for Web Service Discovery, MobiWIS 2012.

Napoli C. Di, Pisa P., and Rossi S., 2013. A market-based

negotiation mechanism for QoS-aware service

selection, The 15th International Workshop on Agent-

Mediated Electronic Commerce.

Kouki J., Chainbi W., and Ghedira K., QoS

Synchronization of Web Services: A Mutli Agent-

Based Model Based Model, 10th International

Symposium on Distributed Computing and Artificial

Intelligence.

CLOSER2014-4thInternationalConferenceonCloudComputingandServicesScience

322