DYNAMIC-AGENTS TO SUPPORT ADAPTABILITY IN P2P

WORKFLOW MANAGEMENT SYSTEMS

A. Aldeeb, K. Crockett and M. J. Stanton on

Department of Computing and Mathmatics

Manchester Metropolitan University, Chester Street, Manchester, M1 5GD, U.K.

Keywords: Dynamic agents, Peer-to-Peer, Exception handling, Workflow adaptability.

Abstract: Peer-to-Peer (P2P) technology is being recognized as a new approach to decentralized workflow

management systems to overcome the limitation of the current centralized Client/Server workflow

management systems. However, the lack of supporting adaptability and exception handling at instance level

of this approach seems to be responsible for the weakness of the P2P workflow management systems.

Dynamic agents can be used within P2P workflow management systems architecture to facilitate

adaptability and exception handling. This paper presents a novel dynamic-agent P2P workflow management

system which integrates three major technologies: software agents, P2P networking and workflow systems.

The adoption of dynamic agents within P2P network can help in overcoming the adaptability problem,

reducing the need for human involvement in exception handling and improves the effectiveness of the P2P

workflow management system.

1 INTRODUCTION

Conventional workflow management systems are

based on centralized client/server architecture. This

requires a centralized database to store the workflow

process definition and a centralized workflow engine

to manage activities such as coordination and

monitoring process execution (Montagut, 2008; Yan

, 2006). The main disadvantages of any such

architecture are the potential bottleneck that can

arise during process execution, and that the central

database can become single point of fault.

In recent years, the integration of workflow and

agent technology has attracted the attention of many

researchers as a means to support distributed

business processes in a dynamic and unpredictable

environment (Mo, 2006; Buhler, 2005; Zhang, 2004;

Müller, 2004; Wang, 2005; Cao, 2005). Agents are

persistent active entities that have the properties of

autonomy, reactivity, and pro-activity and can

perceive, reason, and communicate with other agents

(Singh, 1999). In addition, agents have the capability

to dynamically form social structures through which

they share commitments to the common goal of

workflow enactment by forming a collective entity

called Multi-agent systems (Buhler, 2005). Recent

research has shown an increased interest in P2P

based workflow systems (Aldeeb, 2007; Yan, 2006)

to decentralize workflow systems. P2P is a means of

developing distributed applications where different

nodes, or peers, share resources and have symmetric

roles of either server or client. P2P workflow

management systems (WFMSs) are proposed to

avoid the bottleneck and the central point of fault

caused by centralized client/server workflow

systems. P2P based workflow can also be used to

improve scalability, system openness and support

incompletely specified processes (Yan, 2006). In

P2P based WFMS peers join “virtual communities”

according to their capabilities and discover each

other using the services provided by an open P2P

network. The coordination between peers is

performed by exchange of notification messages.

However, the lack of supporting adaptability and

exception handling at instance level of this approach

seems to be responsible for the weakness of the P2P

WFMS. This paper presents a novel dynamic-agent

P2P WFMS system which integrates three major

technologies: software agents, P2P networking and

workflow systems. The adoption of dynamic agents

within P2P network can help in overcoming the

adaptability problem, reducing the need for human

involvement in exception handling and improves the

effectiveness of the P2P WFMS.

167

Aldeeb A., Crockett K. and J. Stanton M. (2009).

DYNAMIC-AGENTS TO SUPPORT ADAPTABILITY IN P2P WORKFLOW MANAGEMENT SYSTEMS.

In Proceedings of the 11th International Conference on Enterprise Information Systems - Software Agents and Internet Computing, pages 167-170

DOI: 10.5220/0001999101670170

 SciTePress

The rest of this paper is organized as follows.

Section 2 describes the problem and states the

solution adopted. Section 3 describes the proposed

dynamic-agents P2P WFMS. Section 4 illustrates the

adaptability and exception handling process. Section

5 introduces a case study. Finally in section 6, some

concluding remarks are given.

2 PROBLEM STATEMENT

In the centralized Client/Server WFMS the

exceptions are handled and managed by the server.

However, in a P2P WFMS that is fully distributed,

there is no central server and so the exception

handling task poses a major challenge. The main

shortcoming of the proposed P2P WFMS is that it

does not fully support adaptability and exception

handling at instance level. The reason for this

limitation is that peers are bound to a specific

limited set of actions, protocols or messages

specified in the protocols defined for workflow

application (Willmott, 2004). In addition, rational

and cooperative behaviour cannot be guaranteed,

because workflow peers are attached to human

workflow participants. In (Aldeeb et al., 2007) a

mechanism for adaptability and exception handling

in P2P WFMS is proposed. This mechanism is based

on separating the business logic and exception

handling logic using a super-node with exception

handling capabilities (Exception Handling Peer -

EHP) to make it easy to keep track of both. The EHP

captures exceptions, from the workflow peers,

characterises the exceptions and applies a recovery

policy. However, from the initial findings of this

research, extra traffic overhead is generated through

the P2P network due to the P2P conversation

between the Exception Handling Peer and the other

peers which have been affected by an exception.

This can lead to an exponential growth of messaging

load. As a result, the peer’s bandwidth connection

may be exceeded which will affect the overall

system performance. Chen (2000) and Cao (2005)

show that software agents can help in conducting the

run-time function of WFMS, monitoring,

adaptability and exception handling. In addition,

they can move to the appropriate location for high-

bandwidth conversation. This was the motivation of

this research to superimpose the software agents

paradigm on the P2P architecture, where agents can

reside in different peers and perform workflow tasks

utilizing their decision support capabilities, the

collective behaviour, and the interaction protocols of

the P2P system. Furthermore, P2P infrastructure

supports dynamic service construction, modification

and movement, and allows a dynamic agent to

participate in exception handling procedures.

The dynamic agent can be defined as the

software agent that supports dynamic behaviour

modification and does not have fixed set of

predefined functions, however, it carries application-

specific actions, which can be loaded and modified

on the fly (Chen, 2000). Dynamic agents can,

therefore, be used to mediate between the workflow

peers and the exception handling peer to automate

the exception handling process in P2P WFMS. Thus

using dynamic agents within a P2P WFMS can

facilitate system optimization; reducing the

communication overhead, reducing the human

involvement in exception handling and improving

the effectiveness of the P2P workflow management

system.

3 DYNAMIC-AGENT P2P WFMS

A high level architecture of the dynamic-agents

based P2P WFMS is shown in Figure 1. This system

is upgraded from previously designed P2P WFMS

(Aldeeb, 2007). The detailed mechanism and the

functionalities of the peer-to-peer workflow system

can be found in (Yan, 2006; Aldeeb, 2007).

Figure 1: Dynamic-Agent based P2P WFMS.

The new dynamic-agent version of the system is

proposed to take the exception handling function out

of the P2P system level to a dynamic-agent level.

The exception handling function includes: exception

ICEIS 2009 - International Conference on Enterprise Information Systems

168

monitoring, exception detection and exception

handling. The build-time function is conducted at

P2P level which includes workflow process

modelling, storing process definitions and

distributing the process to workflow peers. At run-

time, one of the workflow peers will be responsible

for Instantiating a workflow instance which includes

several tasks. The coordination of tasks will be

achieved by message exchange between peers.

Dynamic agents reside at workflow peers to conduct

exception handling function of the system. As

shown in figure 1, the P2P network provides

services that include advertisement services, group

services, peer services, pipe services, and discovery

services. In addition, it facilitates a user interface

with human workflow participant. The workflow

peers can be classified into three types based on their

capabilities: two super-nodes: Workflow Definition

Peer (WFDP) and Exception Handling Peer (EHP).

The third type is an ordinary Workflow Peer (WFP).

The WFDP facilitates the design and the storage

of the whole workflow schema at build-time. The

workflow process is partitioned to separate tasks

according to the roles of the workflow participants

and the organizational structure. These tasks are

distributed to associated peers.

ExceptionHandlingUserInterface

Exceptions

Data

Dynamic‐Agents

Generationand

Deployment

Caseretrieval

Case

Ada

tation

CaseAnal

sis

Exception

characterisation

EventHandler

CBR



Unit



Figure 2: The workflow exception handling peer.

The EHP has the capability to deal with various

types of exceptions. Figure 2 shows the internal

structure of the exception handling peer. The

presence of this peer is essential for the P2P

workflow management system and if this peer

became unavailable, the system must wait for

another peer with exception handling capabilities to

join the group. The EHP generates and deploys

dynamic agents corresponding to workflow peers.

When an agent is deployed, it registers its symbolic

name and address at its destination node and keeps

an address book for other agents. The dynamic

agents delegate the exception handling functions of

correspondent workflow peers. The exception

handling procedure is discussed in the next section.

The WFP can reside on any machine on the P2P

network enabling direct communication with other

workflow peers to enact the workflow process. Each

WFP performs one task or more of the whole

workflow process. The WFP is associated with a

human workflow participant involved in the

workflow process enactment. Process co-ordination

is achieved by the exchange of messages between

peers.

4 THE ADAPTABILITY

SUPPORT

Workflow adaptability is the ability of the workflow

processes to react to exceptional circumstances

(Sadiq, 2005). Exceptions that occur during

workflow execution have been divided into: basic

failures, application failures, expected exceptions

and unexpected exceptions (Casati, 1998). Basic

failure is related to failures at system level (e.g.,

DBMS, operating systems, or network failure).

Application failure corresponds to failures of any

applications invoked by the WFMS. In the proposed

Dynamic-agent based P2P WFMS, two workflow

exceptions can be identified according to their

location: local workflow exception and global

workflow exception. Local workflow exception

affects the task of one workflow peer. The workflow

peer can handle this exception using its local

dynamic-agent and by applying one of two possible

self-recovery policies; forward recovery or

backward recovery. If the local workflow exception

can not be handled within the affected workflow

peer, its effect may propagate to the other peers

leading to a global workflow exception. These types

of exceptions will, of course, affect more than one

node and a coordinating node is required to deal

with this exception. The coordinating node which

has the capability to deal with the exceptions is the

exception handling peer. This peer generates and

deploys dynamic agents loaded with updated

exception handlers based on prior knowledge. Once

these dynamic agents are deployed to the workflow

peers, they communicate with each other to capture

exceptions, characterize the exceptions and apply a

recovery policy. In addition, these dynamic agents

can be mobile agents which have the capability to

move between the EHP and the exception raising

peers. This can lead to the reduction of the amount

DYNAMIC-AGENTS TO SUPPORT ADAPTABILITY IN P2P WORKFLOW MANAGEMENT SYSTEMS

169

of the communication between nodes as the

interaction will take place locally at the exception

raising node (Cao, 2005). In addition, The EHP will

acquire knowledge from previous exceptions and

updates its exception handling knowledge base to

deal with similar exceptions in future. The EHP is

provided with a Case Based Reasoning (CBR) unit

to handle exceptions which need to be managed in

similar way, but may occur in different instances.

The EHP reloads the dynamic agents with the

acquired exception handling logic in order for these

agents to react to similar exceptions.

5 A CASE STUDY

To better illustrate how the proposed Dynamic-agent

based P2P WFMS works, an example of a motor

insurance claim process is used. The workflow tasks

are distributed over the workflow peers, based on the

roles of the workflow participants and structure of

the organization. After the tasks are distributed, each

workflow peer is aware of its own tasks and the

input and output workflow attributes. The Workflow

tasks in this case study are modelled in Petri Net.

Java has been chosen as the programming language

for this prototype. The P2P network environment is

based on the Java coded Sun Microsystems JXTA.

The workflow peers engines are based on Bossa

workflow engine. Bossa is an open source and

lightweight workflow engine written in Java. In

addition it does not require a RDBMS and is very

simple to use and to integrate with java applications.

For these reasons Bossa is chosen for the prototype

implementation and four Java classes; Place,

Transition, Edge and PetriNet can be adapted for the

P2P environment. To examine the system, build-

time, run-time and exception handling functions are

being implemented.

6 CONCLUSIONS

This paper has proposed a Dynamic-agent based P2P

Workflow Management System. This builds on our

previous work on P2P WFMS. P2P systems and

software agents can fit together for workflow

application. We believe that these two technologies

will play a key role in the future of workflow

systems design and implementation. Dynamic agents

can be used to mediate the workflow peers and the

exception handling peer to automate the exception

handling process. Future work includes: completion

of the different components of prototype and

evaluation of the effectiveness of the system.

REFERENCES

Aldeeb, A., Crockett, K. and Stanton, M. (2007) An

Adaptive P2P Workflow Management System. In

Proceedings of the 9

International Conference on

Enterprise Information Systems, Funchal, Portugal

Montagut, F., Molva, R. and Golega, S. (2008) The

Pervasive Workflow: A Decentralized Workflow

System Supporting Long-Running Transactions. IEEE

Transactions on Systems, Man, and Cybernetics – Part

C, Vol. 38, NO. 3 pp. 319-333

Yan, J., Yang, Y. and Raikundalia, G. (2006) SwinDew- A

P2P-Based Decentralized Workflow Management

System. IEEE Transactions on Systems, Man, and

Cybernetics, 36, No. 5 pp. 922-935

Mo, G and Zhao, W. (2006) Multi-Agent Interaction

Based Collaborative P2P System for Fighting Spam.

In Proceedings of the IEEE/WIC/ACM International

Conference on Intelligent Agent Technology.

Buhler, P. and Vidal, J. (2005) Towards Adaptive

Workflow Enactment Using Multiagent Systems.

Information Technology and Management, 6,pp.61-87

Singh, M., Huhns, M. (1999) Multiagent Systems for

Workflow. International Journal of Intelligent Systems

in Accounting, Finance & Management 8,pp. 105-117

Chen, Q., Dayal, U., Hsu, M. and Griss, M. (2000)

Dynamic-agent, Workflow and XML for E-Commerce

Automation. Lecture Notes in Computer Science, pp.

314-323

Sadiq, S. Orlowska, M. and Sadiq, W. (2005)

Specification and Validation of Process Constraints

for Flexible Workflows. Information Systems, 30, pp.

349-378

Zhang, H. Croft, W., Levine, B. and Lesser, V. (2004) A

Multi-agent Approach for Peer-to-Peer Information

Retrieval Systems. International Conference on

Autonomous Agents and Multi Agent Systems. New

York, USA. ACM,1-58113-864-4/04/0007,pp.456-463

Müller, R., Greiner, U. and Rahm, E. (2004) AgentWork:

A Workflow System Supporting Rule-based

Workflow Adaptation. Data & Knowledge

Engineering 51, pp. 223-256

Wang, M, Wang, H. and Xu, D. (2005) The Design of

Intelligent Workflow Monitoring with Agent

Technology. Knowledge-Based Systems,18,pp.257-

266

Cao, J., Yang, J., Chan, W. and Xu, C. (2005) Exception

Handling in Distributed Workflow Systems Using

Mobile Agents. In Proceedings of the 2005 IEEE

International Conference on e-Business Engineering

Willmott, S., Pujol, J. and Cortes, U. (2004) On Exploiting

Agents Technology in the Design of Peer-to-Peer

Application. Agents and Peer-to-Peer Computing,

LNAI 3601, Third International Workshop, New

York, USA

Casati, F. (1998) Models, Semantics, and Formal Methods

for the Design of Workflows and Their Exceptions.

PhD Thesis, Milan. Italy.

ICEIS 2009 - International Conference on Enterprise Information Systems

170