A P2P-BASED INFRASTRUCTURE FOR VIRTUAL-

ENTERPRISE’S SUPPLY-CHAIN MANAGEMENT

Loris Penserini

, Luca Spalazzi

, Maurizio Panti

ITC-irst, Scientific and Technological Research Institute, Via Sommarive 18, I-38050 Trento-Povo (Italy)

DIIGA, Universià Politecnica delle Marche, Via Brecce Bianche, I-60131, Ancona (Italy)

Keywords: Software Agents and Internet Computing, Artificial Intelligence and Decision Support Systems.

Abstract: This paper proposes and describes a prototy

pe of a peer-to-peer based infrastructure to support virtual

enterprise’s supply chain management. Because of a virtual enterprise is composed of autonomous,

distributed, and continuously evolving entities, we have naturally modelled each business entity like a

peer’s agent platform that can play several roles according to the task to be fulfilled. To this end, we

describe and apply such roles, required to the organizational architecture, into a virtual storehouse scenario.

We thanks D. Iacobucci for suggestions and useful discussions about these topics, M. Orlando and A. Petrini for their

contribution to the implementation phase.

1 MOTIVATIONS

Nowadays, especially in Italy, several organizations,

characterized by common market interests in terms

of products and services they manufacture and

deliver, are collaborating together sharing both their

specific expertise and entrepreneurial culture.

Nevertheless, such a scenario produces a static

terprise coalition always based on the same

members that often have not the competences and

leadership, e.g., in terms of quality, on specific

product and service orders. Furthermore, such

coalitions are generally leader-centered, that is a

coalition’s member (the biggest or the leader one)

establishes and imposes its standards to the other

members.

Actually, several research efforts have been

lfilled in the prospective of endowing small and

medium enterprises with new forms and models of

aggregation and collaboration suitable to take

advantage of current inter-networking information

technology (Franke, 2002; Huhns and Stephens,

2002; Jennings et al., 2000; Pathak et al., 2000;

Petersen et al., 2000).

The scenario introduced above can naturally be

odelled by means of the virtual-enterprise

paradigm (Franke, 2002; Petersen et al., 2000). Into

a virtual-enterprise, each member maintains its

autonomy and independence related to its internal

business processes. Nevertheless, such a member

has to collaborate in a synergic way according to the

coalition goal, e.g., issuing a (new) service and

manufacturing components for a (new) product. This

paradigm establishes a network of small and

medium enterprises characterized by a variety of

value adding products/services (e.g., in a supply

chain), alive only for a specific period, for a specific

business objective, and disband when the goal is

achieved (Franke, 2002).

This paradigm views a distributed system as an

ope

n, dynamic network of peers. Each peer is

acquainted with a small number of other peers with

whom it can exchange information and services.

Acquaintances change constantly, there is no central

control/registry, and peers remain autonomous

throughout their participation in a peer-to-peer (P2P)

network (Bernstain et al., 2002). Notice that, current

P2P tool capabilities are principally based on file

sharing mechanisms; hence, some efforts have been

done in the direction of improving/enhancing the

data management skills, e.g., (Bernstain et al., 2002;

Penserini et al., 2003). For example, in (Bernstain et

al., 2002) the authors proposes an extended

relational model for P2P databases that supports

316

Penserini L., Spalazzi L. and Panti M. (2004).

A P2P-BASED INFRASTRUCTURE FOR VIRTUAL-ENTERPRISE’S SUPPLY-CHAIN MANAGEMENT.

In Proceedings of the Sixth International Conference on Enterprise Information Systems, pages 316-321

DOI: 10.5220/0002612003160321

 SciTePress

distributed query processing and constraint

enforcement.

This paper proposes and describes a prototype of

a P2P framework based on collaborative agents in

order to support and improve the horizontal and

vertical supply chain management network

typologies. Indeed, supply chain management is

considered a strategic and critical aspect for

enterprises and especially for a virtual-enterprise,

e.g., (Huhns and Stephens, 2002; Pathak et al., 2000;

Petersen et al., 2000). In particular, we describe the

ability of the system that can play several roles to

effectively encompass all the supply chain’s

activities, e.g., procurement, production, order

processing, transportation, and customer service.

2 MOTIVATING SCENARIO

The supply chain management is a strategic

component of an enterprise because such an aspect

involves all the activities associated with the value

chain, that is, it copes with the required processes to

transform raw materials to end user products.

Moreover, supply chain management deals with

providing products and services to customer faster,

cheaper, and of better quality, e.g., (Huhns and

Stephens, 2002) and (Petersen et al., 2000).

As an example, it is interesting to observe one of

the main advantages, in terms of costs, in applying a

collaborative model at the chain storehouse

level.

Indeed, generally, the stock holding costs increase

when the product availability increases (i.e., tends to

100%), while the potential lost sale costs increases

when the product availability decreases. Therefore,

the optimal service falls near 85% that is the

minimum of the total cost trend. On the contrary, as

experimented and described in (Netsourcing, 2003),

applying a virtual storehouse collaborative approach

the optimal service moves towards a less percentage

availability, that is less stock investments at the

same customer satisfaction.

For the sake of simplicity, we assume that every

time an order occurs the related enterprise can

satisfy it in three principal ways: a) using its internal

stock, b) negotiating the quantity/lot required with

the known partners, and c) trying to seek for new

partners. Scenario (a), the more traditional one,

means that the firm has to internally produce the

goods required and/or it has to hold a high lot size in

its storehouse in order to satisfy every order. As a

consequence, such an approach risks increasing both

the lost sale and stockholding costs. Those costs are

decreased using scenario (b). That is the known

partners at the same level in the value chain (peer)

contribute to satisfy an (retailer) order. For example,

as depicted in Figure 1, each time a Retailer issues a

sourcing order to Firm

store

Hereafter, by ‘storehouse’ word, we implicitly refer to

the all firm information systems where

data/information are effectively managed and

organized.

, this latter distributes the

order over the known partners, i.e., Firm

relies on

Firm

storehouse. In particular, each firm

autonomously makes their own decisions about how

to bid, e.g., a negotiation phase based on the prices

and their own utilities of the goods

. As a

consequence, such a kind of coalition suffers of

some limitations due to little flexibility to

dynamically reconfigure the enterprise network, and

the (predefined) members have not always the

competence and leadership on specific products or

product’s components. Therefore, such kinds of

alliances tend to adopt common standards that do

not allow other partners to easily come in. Besides,

often in such a setting a central authority (the leader)

constitutes a bottleneck and may break down the

system efficiency completely.

Therefore, in such a context, preliminary

requirements analysis results conduct to small and

medium enterprises that are characterized by weak

technological infrastructure and know-how, e.g.,

they rely upon simple and sporadic Internet

connections; hence, each member can both

autonomously continue to participate inside a

specific market and occasionally exploit

collaboration to increase its buyers’ market. This

latter approach requires being able to know new

partners (scenario (c)). Specifically, this domain can

This paper does not investigate the alternative ways to

fulfill a negotiation.

Figure 1: S im p lified sc e n ario o f v irtu al

house collaboration.

R eta iler

storehouse

Firm

storehouse

Firm

storehouse

Supplier

storehouse

in tern al sto ck

checking

execution

plan (Q ’)

order Q ’

<Q’>

order Q ”

in tern al sto ck

for Q

execution

plan (Q ”=Q

)

<Q”>

order Q

in tern al sto ck

for Q

execution plan

(procu rem en t)

procurem ent

order

stock restoring

Figure 1: Simplified scenario of virtual storehouse

collaboration

A P2P-BASED INFRASTRUCTURE FOR VIRTUAL-ENTERPRISE'S SUPPLY-CHAIN MANAGEMENT

317

be accommodated in a natural way by means of a

P2P network. Thus, this style of computing is very

suitable for collaborative actor groups, e.g., virtual-

enterprise’s members that work under conditions of

autonomy, coordination, and not permanent

connections. In this paper, we aim to extend and

improve the scenario (b) and (c) endowing each

coalition member with an agent based peer-to-peer

(P2P) framework.

3 APPROACH

The proposed multi-agent system

aims to

characterize the principal agent roles and their

relationships required to support and enhance

information coordination in a virtual-enterprise’s

supply chain scenario. Namely, the proposed system

does not aim to substitute the internal enterprise

behaviour and features, e.g., logistics, supply chain

management system, and information systems. On

the contrary, it allows enterprises for supporting a

more dynamic P2P based computing approach to

model the collaborative interactions among partners.

In particular, we dedicate more focus on virtual-

enterprise composed of autonomous members

(peers) with fragmentary information about the

environment in which they live, e.g., incomplete

information on both partners and business processes;

hence, they exploit coordination each other in order

to achieve common goals. For such reasons, our

multi-agent system (the peer’s agent platform)

supports the peer-to-peer computing model.

Moreover, we assume that each peer node includes a

peer (the enterprise) and a (software) peer’s agent

platform, which assists the peer (see Figure 2).

As indicated in Figure 2, each agent platform

deploys one or more of the following capabilities to

support the needs of its human/organizational peer:

Facilitator (searching and registration). In the

scenario (b) and (c) described in Section 2, each peer

(i.e., an enterprise) needs of looking for partners

capable to satisfy a given request. That is, a virtual-

enterprise can be seen as a decentralized agent

setting, in which each peer (a virtual-enterprise’s

member) does not know a priori what partners to

communicate with. Therefore, the facilitator role

allows the peer’s agent platform for the searching

and registration capabilities in order to get to know

other peer’s agent platforms with useful skills,

establishing new acquaintances. For example, in our

approach this ability is based on a yellow page

directory service, where agents can advertise

themselves and their functionalities. Yellow pages

also provide information about the state (e.g., active,

disconnected) of other agents and platforms, e.g., see

(Penserini et al., 2002; Fipa, 2000). Moreover, as

depicted in Figure 2, each time a request cannot be

internally satisfied, both the supply chain manager

and the information source manager could interact

with the facilitator role to get new acquaintances,

that is the scenario (b). Notice that, in the case of a

new peer’s request, the facilitator can also

autonomously propagate the request over the peer

network without overloading the supply chain

manager, e.g., interacting with other platforms’

facilitators, that is the scenario (c). In our prototype,

we adopt the ‘Facilitator’ name, because such an

agent has been originally developed according to the

Fipa’97 reference model (Fipa, 2000), but its further

functionality improvements now make it similar to a

matchmaker agent role.

It is partially based on a MAS prototype, named JEAP.

Available at: http://jeap.inform.unian.it/.

Information System Manager (reformulation

and integration). When a given peer (enterprise)

operates in the scenarios (b) and (c) of Section 2, it

needs to interact with the information systems of

other peers; this is a sort of virtual storehouse. In

other words, we are in the presence of a distributed

and heterogeneous information systems. In

particular, each peer’s agent platform relies on this

role to perform and coordinate queries targeted to

information sources of the same peer or other peers.

Therefore, there exists a well-known data integration

problem in distributed, heterogeneous, and dynamic

systems. Hence, to cope with integration issues, the

peer’s agent platform can adopt a mediator

architecture to access the information sources, e.g.,

in our prototype, this architecture is composed of

mediator and wrapper agents. For example, it can

use one of the several algorithms for answering

queries using views, e.g., see (Panti et al., 2001).

Therefore, the information system manager (is

facilitator

manager

5: ask for

collaboration

3: ask for

partner

3: stock

status

2: ask for

stock

info

1: initial

request

’s agent platform

facilitator

manager

’s agent platform

7: motivated

answer

4: partner

list

Figure 2:

Principal roles played by a peer’s

t platform. agen

Figure 2: Principal roles played by a peer’s agent

platform

ICEIS 2004 - SOFTWARE AGENTS AND INTERNET COMPUTING

318

manager) provides a platform with reformulation

and integration capabilities. Using these, a platform

can reformulate the original problem (initial request)

in terms of data management operations each

targeted at selected sources, in agreement with some

soft inter-database constraints, i.e., coordination

rules as described in (Bernstain et al., 2002).

Supply Chain Manager (strategy generation).

The supply chain manager (sc manager) role is

required to correctly coordinate collaboration

activities among decentralized peers, i.e., virtual-

enterprise members. For instance, when a failure

results from the peer’s agent platform inability to

satisfy a request locally, the supply chain manager

can help to build up a cooperation strategy in order

to overcome the underling failure. In particular, our

system prototype’s supply chain manager currently

deals with the principal failures that affect virtual

storehouse scenario, such as: product stock

unavailability and procurement criteria, partners’

unavailability and looking for new partners.

Specifically, the supply chain manager manages

plans (workflows) composed of actions in order to

fulfil negotiations, to query information sources, to

make bids, etc. Indeed, in our preliminary tests as

shown in Figure 2, we have assumed that the pivot

role is played by the supply chain manager role; that

interprets and manages every initial request to

choose the more convenient plan. To this end, such a

role can rely on the well-known BDI (Belief-Desire-

Intention) architecture. According to this

architecture, the supply chain manager represents

the environment status in terms of facts (the beliefs)

and the received requests in terms of goals (the

desires). Moreover, it chooses the more convenient

behaviour (the intentions), among a set of plans, to

achieve the current goal. Finally, each supply chain

manager has the responsibility of coordinating the

internal activity required to keep update all the

enterprise’s repositories. The supply chain manager

relies on the agent platform’s information system

manager in order to inquire the peer’s internal

information sources (required to update its beliefs),

e.g., repositories to get stock status about specific

products.

4 INTERACTION EXAMPLES

Let us assume that the entities shown in Figure 1,

i.e., Retailer (R), Firm

), Firm

), and

Supplier (S) are respectively interfaced by their

platforms (peer’s agent platform), i.e., PA

, PA

, and PA

Specifically, our prototype support an agent

communication language (ACL), based on the FIPA

specification (Fipa, 2000), in order to standardize

and define the format of the exchanged messages.

ACL is fundamental in order to allow agents to

understand their intentions. Besides, performatives,

e.g., List 1 shows the ‘request’, are constrained to

follow an exact path of the discourse, required to the

agents to recognize if a conversation is terminated

(and for what reason) or if it is still in progress, in

which point it is (say communication protocol).

To make clearer the roles played by each

platform, let us assume to fulfil the order Q”

indicated in Figure 1.

(request

:sender retailer

:receiver PA

(sc)

:language XML

:ontology planner-strategy

:protocol fipa-request

:content

<action> PA

(sc) </action>

<ID_product> Acer_LCD17_01 </ID_product>

</availability>

:conversation-id #)

Table 1: An example of ‘request’ message.

re 3: Cooperation plan to cope with the

o (b) depicted in Figure 1

Figu

scenari

(sc)

stock status

checking

info on Q”

<Q”>

ask for Q

(is) PA

(sc) PA

(is)

order Q”

stock status

checking

info

lan

selecting

ask for Q

info on Q

stock status

checking

info

lan

selecting

rocurement

orde

ask for Q

’s agent platform

Figure 3: Cooperation plan to cope with the

scenario (b) depicted in Figure 1

A P2P-BASED INFRASTRUCTURE FOR VIRTUAL-ENTERPRISE'S SUPPLY-CHAIN MANAGEMENT

319

Example 1. According to Figure 3, the order Q”

is submitted to the supply chain manager role of

Firm

(PA

(sc)) by means of a ‘request’ message,

i.e., indicating the product ID and the requested

quantity (Q”) as shown in List 1. Consequently,

(sc) needs to extract all the product information

to effectively deal with the choice of the more

convenient execution plan; hence, it relies on the

information system manager role (PA

(is)). As

already said, PA

(is) is able both to reformulate the

supply chain manager’s request into information

source requests, and, vice versa, to integrate the

source’s answers into a single and coherent answer

to the supply chain manager. For instance, in List 2

is shown the ‘inform’ message content that PA

(is)

provides to PA

(sc), that is the ‘info’ message line

of Figure 3. Therefore, when PA

(sc) receives the

product info, e.g., stock status, procurement criteria

(as the Pareto’s law), etc.; it has the main

components to select the more convenient execution

plan. As already said, in this preliminary tests, our

prototype’s supply chain manager aims to avoid

internal procurement order collaborating with other

partners in the same level of the value chain. To this

end, according to the ‘inform’ message of List 2,

(sc) splits the order Q” in two sub-orders: the

first one (with quantity Q

) satisfied by the Firm

storehouse and the second one (with quantity Q

)

delegated to Firm

storehouse. Notice that, Firm

also involved in a stock restoring phase, i.e.,

procurement order.

Figure 4:

A fragment of the cooperation

strategy to cope with scenario (c)

(sc)

acquaintances

updating

…

(f) PA

(f)

order Q

choose

strategy

new peer

list

…

ask for peers

ask for new peers

eer list

list

choose

strategy

negotiating phase

. . .

Example 2. Let us complicate a bit the scenario

described in Figure 3. Namely, the order Q” consist

of a new product request (quantity Q”) and Firm

does not know partners, among its current

acquaintances, able to collaborate on such a request.

Consequently, Firm

is forced to seek for new

partners (peers) in order to avoid the lost sale costs

increasing. Notice that, such a case coincides with

the scenario (c) introduced in Section 2.

Figure 4: A fragment of the cooperation strategy to

cope with scenario (b)

Figure 4 depicts the scenario (c) in terms of the

main involved interactions among peers’ agent

platforms and their roles. When PA

(sc) realizes

that it cannot satisfy the order alone, it relies on the

facilitator role (PA

(f)) to find new partners. The

alternative ways to fulfil collaboration (cooperation

plan) drive us to the notion of cooperation

strategies

. Therefore, Figure 4 indicates such a

process by the ‘choose strategy’ label. In order to

satisfy the request ‘ask for new peers’, PA

(f)

performs a broadcast request over its local

acquaintances (peers). In particular, by the facilitator

role, peer agent’s local acquaintances enable access

to global information; namely, each peer’s global

behaviour emerges from local interactions, e.g., see

(Penserini et al., 2003; Penserini, 2002).

….

:content

<session> SUCCESSFUL </session>

<ID_product> Acer_LCD17_01 </ID_product>

<safety-level> 20 </safety-level>

<flag_availability> on </flag_availability>

<ID_warehouse> Firm

_warehouse3

</ID_warehouse>

</Element>

</SELECT1> </selectresults>

</xmlcontent>

….

Table 2: Fragment of an ‘inform’ message generated by

an information system manager role

In our simple tests, for the strategy component

‘ask for new peers’, we decided to select only the

active supply chain managers of each peer’s agent

platform. Moreover, the PA

(sc) has to establish a

collaborative criteria (negotiating phase) in order to

effectively fulfil the order Q. For the sake of

simplicity, such a supply chain manager organizes

its partners on a product availability basis, that is, it

Despite of a cooperation strategy is composed of several

components (Panti et al., 2001; Penserini et al., 2003),

in such an example, we only describe the way to fulfil

the ‘ask for new peers’ component.

ICEIS 2004 - SOFTWARE AGENTS AND INTERNET COMPUTING

320

fairly distributes the requested quantities trying to

avoid to each partner a procurement phase. Each

supply chain manager adopts the well known

crossed matrix (or ABC) criteria based on the

Pareto’s law in order to decide the optimal

procurement point, e.g., List 2 shows that the Acer

LCD display belongs to the class AA of the crossed-

matrix.

5 RELATED WORK AND

CONCLUSIONS

Recent contributions to the VE’s supply chain

modelling come from the promising area of multi-

agent systems, e.g., (Jennings et al., 2000; Pathak et

al., 2000; Petersen et al., 2000).

In (Pathak et al., 2000) the authors describe how

to model a supply chain scenario exploiting both the

ZEUS Agent Building toolkit and the Generic

Modeling Environment (GME). Such tools are able

to deal with the problem analysis, the problem

design, and the application realization phases.

Moreover, they explain how to use the tools to

define the agent properties (e.g., ontology, tasks,

communication protocols) focusing more on agent

modelling issues rather than on the agent

coordination. Despite of they recognize the

remarkable aspect that agents are often not perfectly

rational and fully informed about the world in which

they work; they do not characterize any agent roles

to overcome such limitations. The AGORA multi

agent architecture, described in (Petersen et al.,

2000), aims to model functional aspects and the life-

cycle of a virtual enterprise. The authors consider a

homogeneous environment composed of enterprises,

which use the same AGORA system, able to form

temporary coalitions (VEs). That is, they do not deal

with enterprise information systems heterogeneity.

Furthermore, such a system can not model a peer-to-

peer scenario characterized by autonomous entities

(peers) free to interact each other without to a priori

build up a central authority, i.e., an AGORA’s

instance, that control their communications.

Probably one of the most complete and flexible

agent-based approaches to model and build process

management systems has been realized inside the

ADEPT project (Jennings et al., 2000). For instance,

ADEPT allows designers for dynamic, distributed,

and unpredictable processes, besides, it is able to

manage multiple (even though decentralized)

organizations that concurrently participate to a given

process. The authors detail their system architecture

and functionalities, but no examples are given about

how to configure each agent’s role into a real world

scenario, e.g., VE’s supply chain scenario.

Thus, our paper presents and describes an agent-

based framework prototype, required to cope as

automatically as possible with virtual enterprise’s

supply chain management coordination issues. That

is, each peer (person/organization) has an (software)

agent platform that manages the peer’s participation

in the P2P coalition (virtual-enterprise). As far as we

know, the results of the architectural design analysis

in terms of agents’ roles, needed to effectively deal

with the inherently interaction relationships of a

supply chain scenario, is quite original.

REFERENCES

Bernstein, P., Giunchiglia, F., Kementsietsidis, A.,

Mylopoulos, J., Serafini, L., Zaihrayeu, I., 2002. Data

Management for Peer-to-Peer Computing: A Visions.

In Proc. of the Fifth International Workshop on the

Web and Databases (WebDB).

Fipa, 2000. Fipa. The foundation for intelligent physical

agents: FIPA 2000 specification (www.fipa.org).

Franke, U. J., 2002. Managing Virtual Web Organizations

in the 21th century: Issues and Challenges. Idea

Group Publishing, Pennsylvania.

Huhns, M. N., Stephens, L. M., 2002. Automating Supply-

Chain Management. In Proc. of the 1

ACM Joint

Conference on Autonomous Agents & Multiagent

Systems, pp.1017-1024, ACM Press, Bologna, Italy.

Jennings, N. R., Norman, T. J., Faratin, P., O’Brien, P. and

Odgers, B., 2000. Autonomous Agents for Business

Process Management. Int. Journal of Applied Artificial

Intelligence, 14(2), 145-189.

Netsourcing, 2003. UNITEC GmbH. Available at:

http://www.netsourcing.it/

Pathak, S. D., Nordstrom, G., Kurokawa, S., 2000.

Modeling of supply chain: a multi-agent approach. In

Proc. of 2000 IEEE International Conference on

Systems, Man, and Cybernetics, vol. 3, pp. 2051-2056.

Panti, M., Spalazzi, L., Penserini, L., 2001. Cooperation

Strategies for Information Integration. In Proc. of the

Sixth International Conference on Cooperative

Information Systems (CoopIS), Springer Verlag.

Penserini, L., Lin, L., Mylopoulos, J., Panti, M., Spalazzi,

L., 2003. Cooperation Strategies for Agent-Based P2P

Systems. WIAS: Web Intelligence and Agent Systems:

An International Journal, Publisher: IOS Press.

Penserini, L., 2002. Integration and Coordination in both

Mediator-Based and P2P Systems. Ph.D. Thesis,

Istituto di Informatica, University of Ancona, Italy.

Petersen, S. A., Divitini, M., Matskin, M., 2002. An

Agent-based Approach to Modelling Virtual

Enterprises. International Journal of Production

Planning and Control, 2000.

A P2P-BASED INFRASTRUCTURE FOR VIRTUAL-ENTERPRISE'S SUPPLY-CHAIN MANAGEMENT

321