Using Petri Nets to Enable the Simulation of

Application Integration Solutions Conceptual Models

Fabricia Roos-Frantz

, Manuel Binelo

, Rafael Z. Frantz

, Sandro Sawicki

and Vitor Basto-Fernandes

UNIJUÍ University, Department of Exact Sciences and Engineering, Ijuí, Brazil

Polytechnic Institute of Leiria, School of Technology and Management, Leiria, Portugal

Keywords:

Simulation, Stochastic Petri Net, Enterprise Application Integration, Integration Solution, Integration Patterns.

Abstract:

Enterprise application integration concerns with the use of methodologies and tools to design and imple-

ment integration solutions to integrate a set of heterogeneous enterprise applications. Amongst the integration

technologies to design and implement integration solutions is Guaraná. This technology provides a domain-

speciﬁc language that enables the design of conceptual models. The quality of these models is essential to

ensure proper integration. Discovering whether an integration solution can fail and in which conditions it is

more likely to happen is a costly, risky, and time-consuming task, since current approaches require software

engineers to construct the real solution. Generally, simulation is recommended when problems are impossible

or expensive to be solved by actual experimentation. Guaraná conceptual models can be classiﬁed as stochas-

tic, dynamic, and discrete, and thus it can be simulated taking the advantage of well-established techniques

and tools for discrete-event simulation. Therefore, this paper proposes a simulation of Guaraná solutions by

using Petri nets, in order to analyse such solutions based only on the conceptual models. It shows that an inte-

gration solution conceptual model designed with Guaraná can be translated into a formal model structured as

a Stochastic Petri net. The equivalence of both models is veriﬁed by comparing the operation of the Guaraná

runtime system with the behaviour of a Petri net execution process.

1 INTRODUCTION

Today’s companies face the challenge of integrating

their usually heterogeneous software ecosystem and

reduce the high costs associated with this integration.

Companies usually spend nearly 5 to 20 times more

to integrate their applications than to develop a new

functionality (Weiss, 2005). To cope with this prob-

lem, companies rely on methodologies and tools to

design and implement integration solutions provided

by the Enterprise Application Integration (EAI) com-

munity. An integration solution is able to keep a num-

ber of applications’ data in synchrony or to develop

new functionality on top of them, without changing

or disturbing the applications being integrated (Hohpe

and Woolf, 2003).

Currently, there are a number of integration tech-

nologies to design and implement integration solu-

tions (Ibsen and Anstey, 2010; Fisher et al., 2010;

Dossot and D’Emic, 2009; Frantz and Corchuelo,

2012). The EAI community is pushing the integra-

tion technologies to provide support for the realisa-

tion of the enterprise integration patterns documented

by Hohpe and Woolf (2003). These conceptual pat-

terns document a set of well-known services that are

directly related to integration tasks and have been

turned into a cookbook for software engineers to de-

velop their solutions. The design of conceptual mod-

els is usually supported by a domain-speciﬁc lan-

guage built on top of these patterns. In this paper we

focus on Guaraná integration technology (Frantz and

Corchuelo, 2012). Guaraná not only provides sup-

port for the design of conceptual models using such

kind of domain-speciﬁc language, but it also proposes

a model-centric development approach. According

to the authors, Guaraná is based on the design of

platform-independent models that can be reﬁned, by

means of model transformation techniques, to gener-

ate the corresponding source code for the integration

solution.

The quality of conceptual models for integration

solutions is essential to ensure proper integration.

Discovering whether an integration solution can fail

and in which conditions it is more likely to happen

Roos-Frantz F., Binelo M., Z. Frantz R., Sawicki S. and Basto-Fernandes V..

Using Petri Nets to Enable the Simulation of Application Integration Solutions Conceptual Models.

DOI: 10.5220/0005351900870096

In Proceedings of the 17th International Conference on Enterprise Information Systems (ICEIS-2015), pages 87-96

ISBN: 978-989-758-096-3

 2015 SCITEPRESS (Science and Technology Publications, Lda.)

is a costly, risky, and time-consuming task. It incurs

because the current approach practiced by software

engineers requires the construction of the complete

integration solution, its execution in a runtime sys-

tem, and the collection of data from this execution,

so that the integration solution can be analysed. An

approach that enables the analysis of the integration

solution without having to construct it, taking as in-

put only the conceptual models devised at the design

phase, would help to reduce cost, risk, and time spent.

A conceptual model is characterised by its high

level of abstraction. A good solution would be

to translate a conceptual model into another high

level abstraction model that relies on automated

analysis support, since the goal is not to generate

source code. In the Model-Driven Engineering dis-

cipline (Schmidt, 2006), such transformation is clas-

siﬁed as horizontal and exogenous transformation. “A

horizontal transformation is a transformation in which

the source and target models reside at the same ab-

straction level” and “exogenous transformations are

transformations between models expressed using dif-

ferent languages” (Mens and Gorp, 2006).

Guaraná conceptual models can be classiﬁed as

stochastic, dynamic, and discrete. As a discrete sys-

tem, it can be simulated taking the advantage of well-

established techniques and tools for discrete-event

simulation. Simulation models are used to get an-

swers about the operation of the system and may

be able to analyse several aspects of it. Therefore,

to be able to analyse Guaraná solutions based only

on the conceptual models, we propose a simulation

of such solutions by using Petri nets (Bause and

Kritzinger, 1998). This mathematical modelling lan-

guage can be used to model discrete event systems

of any kind (David and Alla, 1994), and their graph-

ical nature allows the visualisation of the system be-

haviour comprising concurrency, synchronisation and

resource sharing. They are executable, have a math-

ematically deﬁned semantics, and provide a mathe-

matical theory for analysis and veriﬁcation of certain

properties of a process.

Guaraná follows the Pipes and Filters architectural

style (Hohpe and Woolf, 2003). In this style, a larger

process is divided into a number of smaller and inde-

pendent services (Filters), which are usually unsyn-

chronised by channels (Pipes). In an integration pro-

cess, there are two main components: i) pipes, imple-

mented by slots, and ii) ﬁlters, implemented by tasks.

In an integration solution, messages ﬂow through in-

tegration processes, going from slots to tasks or from

tasks to slots, and are processed asynchronously. Sim-

ilarly, Petri nets are structured as follows: i) they are

bipartite graphs, since they have two types of nodes,

places and transitions, ii) they are directed graphs,

arcs can connect places to transitions and transitions

to places only, and iii) they are multi-graphs, its pos-

sible to have multiple parallel arcs between a place

and a transition. As in the case of an integration so-

lution, in which the execution of a task depends on

the availability of messages in all slots connected to

its inputs, in Petri nets the execution of a given transi-

tion occurs only when a token is available in all input

places connected to such transition.

The aim of this paper is to show that an integration

solution conceptual model designed with Guaraná can

be translated into a formal model structured as a

Stochastic Petri net. We propose to verify the equiv-

alence of both models by comparing the operation of

the Guaraná runtime system with the behaviour of a

Petri net execution process.

According to Rezai et al. (1995), Petri nets have

proven themselves to be an excellent modelling and

analysis tool for discrete event or asynchronous sys-

tems. In this work, the authors use a new extension

of Petri nets, called Global Petri net that provides a

means for combining differential and differenceequa-

tions to represent time dependent or synchronous sys-

tems. The goal is to show how it can be used to model

and simulate hybrid systems. An example of real time

control systems was used to show the Global Petri net

modelling and simulation capabilities.

Zhou (1998) presents a Petri net approach aim-

ing at modelling, analysing, simulating, scheduling,

and controlling for semiconductor manufacturing sys-

tems. This paper presents the fundamental con-

cepts of Petri nets, its importance for qualitative and

quantitative analysis of systems. The author also

describes that this kind of system can be charac-

terised as discrete-event systems, however, it can ex-

hibit sequential, concurrent, and conﬂicting relations

amongst the events and operations having its evolu-

tion dynamic over time.

Cavalieri (2000) proposes a methodology to im-

prove the performance of a Flexible Manufacturing

System (FMS), based on the use of Petri nets. The

goal is to model the system to represent the general

behaviour of the FMS, avoiding a detailed description

of its operations. The Petri net model of the FMS is

used to evaluate its performance through simulation.

Kuo (2004) proposes a simulation and control

environment for the distributed event dynamic sys-

tems, based on the distributed agent-oriented Petri

net, a type of high-level Petri nets. Davidrajuh (2008)

presents a new Petri net tool for modelling and simu-

lation of discrete-event systems. According to Alla

and Ghomri (2012), Petri nets are widely used to

model discrete event dynamic systems. The paper,

ICEIS2015-17thInternationalConferenceonEnterpriseInformationSystems

the authors have presented the basic ideas describing

the continuous and hybrid models. In addition, they

describe that continuous and hybrid Petri nets allow

modelling and analysis of continuous and hybrid sys-

tems on the same conceptual basis.

Narahari et al. (1989) report that the Stochas-

tic Petri nets have emerged as a main performance-

modelling tool for distributed systems and the

discrete-event simulation is the preferred tool for its

performance evaluation. The authors present several

algorithms based on Stochastic Petri nets aiming at

discrete-event simulations of distributed systems.

David and Alla (1994) present the basic concept

of Petri nets and the classes of derived models that

can be used for dynamic system modelling. The au-

thors report that autonomous Petri nets enable a dis-

crete event system of any kind to be modelled. Then,

the relationship between system and its model based

on Petri nets allows the qualitative validation of the

functioning process.

Gold (2004) presents an integrated software engi-

neering approach for the use of Petri nets in software

development from analysis to testing of software. The

author uses the classical software development pro-

cess with simulation and mathematical analysis. The

main advantage is that requirements can be validated

earlier, thereby, the fault detection and correction is

less expensive. In addition, the results can be simu-

lated and analysed using Petri net tools.

Other authors also propose the use of Petri nets for

conceptual modelling. van der Aalst et al. (2013) pro-

pose the use of Coloured Petri nets to model complex

process, arguing that this extension of a classical Petri

formalism is more expressive and thus it is not neces-

sary to abstract from relevant aspects of such complex

process. Russell et al. (2009) also apply Petri net for-

malism to get modelling beneﬁts since they present

commonalities with the domain of workﬂow systems.

From the analysis of the literature, it is clear that

Petri nets have been widely used as support for sys-

tem modelling and simulation, nevertheless it has not

been explored yet in the context of EAI to simulate

integration solutions taking as input their conceptual

models.

The rest of this paper is organised as follows: Sec-

tion 2 gives an overview of Guaraná technology and

Stochastic Petri nets; Section 3, presents our proposal,

in which we show how to translate Guaraná concep-

tual models into Petri net formal models; Section 4

presents a case study; Section 5 discusses the analy-

sis results obtained from the simulation of the formal

model, and, ﬁnally, Section 6 concludes this paper.

2 BACKGROUND

In this section, we provide a brief overview of

Guaraná technology and Stochastic Petri nets.

2.1 Guaraná

Guaraná technology is the result of a research effort

to provide support for the design, implementation,

and execution of enterprise application integration so-

lutions. This technology is composed of a domain

speciﬁc language, a software development kit, and a

runtime system. The domain speciﬁc language is a

graphical modelling language used to support soft-

ware engineers in the design of integration solutions

at a high level of abstraction (i.e., conceptual models)

and based on the well-known integration patterns doc-

umented by Hohpe and Woolf (2003). The software

development kit provides a Java application program-

ming interface to implement conceptual models into

executable code. The runtime system is used to exe-

cute the implemented integration solutions.

Guaraná language has a number of constructors,

each of them representing a given application integra-

tion concept. Messages are an abstraction of a piece

of information that are exchanged and transformed

across an integration solution. They are composed of

a header, a body, and one or more attachments. Tasks

represent an atomic operation that can be executed on

messages. They model how a set of inbound messages

must be processed to produce a set of outbound mes-

sages, such as routing the inbound messages, modify-

ing them, transforming them, performing time-related

actions, and mapping them to/from objects. Slots are

buffers connecting an input of a task to the output of

another task, providing asynchronous message pro-

cessing. Ports abstract processes away from the com-

munication mechanism in an inter-process communi-

cation or in the communication of the integration so-

lution with an application. Processes are the central

processing units in an integration solution, they are

composed of ports and tasks.

state 1

state 2

Legend

task

slot

message

Figure 1: Sample of task execution.

In a conceptual model, an integration solu-

tion represents one or more integration processes

UsingPetriNetstoEnabletheSimulationofApplicationIntegrationSolutionsConceptualModels

trough which messages ﬂow and are processed asyn-

chronously. The workﬂow is implemented as a Pipes

and Filters architecture, in which pipes are imple-

mented by Slots and ﬁlters are implemented by Tasks.

Tasks become ready to be executed every time there is

a message available in every input. Becoming ready

for execution implies that the runtime system creates

a work unit to register the task into a queue of exe-

cutable tasks, and latter can assign a thread to execute

it; however, this does not entail that the task produces

a set of outbound messages, since it depends on the

inbound messages and the semantics implemented by

the task. For instance, a ﬁlter is a task that reads a

message from its input and may or may not remove

this message from the workﬂow; such a task eventu-

ally may not produce outbound message. As slots en-

able asynchrony in an integration solution, messages

are stored on them until the tasks are able to process

these messages. When a task is executed, the inte-

gration solution changes its state, this concept is il-

lustrated in Figure 1. A detailed discussion on the

domain-speciﬁc language of Guaraná is provided by

the authors in Frantz et al. (2011).

2.2 Stochastic Petri Net

In 1962 Carl Adam Petri invented the Petri nets as a

formalism to describe the concurrency and synchro-

nisation dynamics in distributed systems (Bause and

Kritzinger, 1998). Petri nets are graphs composed of

two kinds of nodes, places, and transitions. Places are

represented by circles and Transitions by rectangles,

these nodes are connected by arcs. Places can contain

tokens, represented by dots inside the place, and the

amount of tokens in a place is called a marking. An

input arc connects a place to a transition and an output

arc connects a transition to a place. When a transition

is ﬁred, the tokens in the places connected to input

arcs are removed and tokens are added to the places

connected to the output arcs. A transition is active

and can be ﬁred if the amount of tokens determined

by the input arcs exists in the respective places. The

amount of tokens generated by the output arcs are not

necessarily the same as the removed by the input arcs.

When a transition ﬁres, the Petri net changes its state,

this concept is illustrated in Figure 2.

Various variations of Petri nets have been used to

model different classes of problems. Amongst them,

the continuous-time stochastic Petri net, or simply

stochastic Petri net (Molloy, 1981), formed by the tu-

ple (P,T,I, O,M

,Λ), where P is the set of places, T is

the set of transitions, I is the set of input arcs, O is the

set of output arcs, M

is the initial marking and Λ is

the set of transition rates. The transition rate is a ﬁring

state 1

state 2

Legend

place

transition

token

arc

Figure 2: Sample of ﬁring a transition.

delay for the transition determined by a random vari-

able X with a negative exponential probability density

function F

(x) = 1− e

−λ

where − λ

is the ﬁring rate

of transition t

. The ﬁring rate is marking dependent

and the average ﬁring delay of transition t

in marking

is [λ

)]

−1

(Marsan, 1990).

3 PROPOSAL

In this work, we propose the translation of the

Guaraná model into a stochastic Petri net model. This

translation brings several advantages. The ﬁrst beneﬁt

is the fact that stochastic Petri nets are a very well es-

tablished mathematical model, they can be translated

into other formal models and provide various statis-

tical analysis of the model behaviour. Another ben-

eﬁt is that Stochastic Petri have simple components

(places, transitions, and arcs), are very generalisable,

and there are numerous tools available for Petri nets

simulation and analysis.

In order to translate the Guaraná model into a Petri

net, it is necessary to use a table, our Rosetta Stone,

to translate each Guaraná component into an equiva-

lent Petri net. The translation of a speciﬁc Guaraná

solution into a stochastic Petri net is carried out by

connecting these Petri net components in the same

fashion of the components in the Guaraná solution.

In this paper we have considered the following kinds

of tasks: communicator, mapper, modiﬁer, router,

stream dealer, and timer. Communicator tasks enable

to send or receive messages through ports to applica-

tions being integrated. Mapper tasks allow to map the

body of a message from XML to Stream or vice-versa.

Modiﬁer tasks add or remove data from the body and

the head of inbound messages. Router tasks get one

or more messages from its input slots and transfer or

copy them to one or more output slots. Stream dealer

tasks allow to operate on messages whose body con-

tains raw data. Timer tasks allow to execute timer-

ICEIS2015-17thInternationalConferenceonEnterpriseInformationSystems

GuaranáPetri net Task name

In Communicator

Out Communicator

In-Out Communicator

Out-In Communicator

Figure 3: Communicator tasks.

XML2Stream

Stream2XML

GuaranáPetri net Task name

Figure 4: Mapper tasks.

GuaranáPetri net Task name

Promoter

Header Demoter

Header Enricher

Context-based Slimmer

Content Enricher

Context-based Content Enricher

Context-based Header Enricher

Set Correlation ID

Set Return Address

Figure 5: Modiﬁer tasks.

based operation on messages. Figures 3, 4, 5, 6, 7, 8

and 9 show the translation of each type of task from

Guaraná into its Petri net representation.

The translation from Guaraná to Petri net is a one

way process. This happens because, even though

all information necessary to the time and concur-

rency simulation can be translated to the Petri net,

the domain-speciﬁc language task components have

semantic information that cannot be translated. The

Corr

Merger

Resequencer

Filter

Idempotent Transfer

Distributor

Replicator

Dispatcher

Semantic Validator

Threader

Guaraná

Petri net Task name

Figure 6: Router tasks.

Zipper

Unzipper

Encrypter

Decrypter

Encoder

Decoder

Guaraná

Petri net Task name

Figure 7: Stream dealer tasks.

result is that various semantically different Guaraná

components are mapped to identical Petri net. Take

for example the components zipper and content en-

richer, both have different meaning in the domain-

speciﬁc language of Guaraná, but share the same ab-

stract functionality, that take a message, do an opera-

tion on the message and output the message. Both are

represented by the Petri net composed of two places

and one transition, as can be seen in Figures 5 and 7.

When we translate a solution, every message in

Guaraná becomes a token, and every work unit in

Guaraná becomes a transition ﬁring in Petri net, so

that the execution of the stochastic Petri net is equiv-

alent to the execution of the single-threaded Guaraná

runtime system.

UsingPetriNetstoEnabletheSimulationofApplicationIntegrationSolutionsConceptualModels

Delayer

Expire Checker

Guaraná

Petri net Task name

Figure 8: Timer tasks.

Aggregator

Splitter

Translator

Assembler

Chopper

Cross Builder

GuaranáPetri net Task name

Figure 9: Transformer tasks.

The ﬁring probability, or ﬁring rate of each tran-

sition of the Petri net, may have two different mean-

ings in the translated model. First, it can be the ac-

tual ﬁring rate, i.e., how many times an operation was

performed. One example is the ﬁlter in Figure 10,

with transition T

and T

with the respective ﬁring

rates a and b. Transition T

lets messages (tokens)

pass to the next stages, while transition T

removes

messages from the system. If rate a = 0.9 and rate

b = 0.1, then 10% of messages are ﬁltered out from

the system. Second, it can be the time a task takes

to execute. This interpretation can be made because

stochastic Petri nets (also called temporal Petri nets),

can be analysed from a temporal perspective. The

probability associated with the transition determines

its frequency of ﬁring, and since time is the inverse

of frequency, the time consumed for each task can

be computed by t = 1/ f, in work unities. Using a

slimmer task as example, if the transition has a ﬁre

rate of 0.5, which means that the task has an average

processing time of two work unities.

T (b)

T (a)

Figure 10: Petri net representation for a ﬁlter task.

4 CASE STUDY

This case study consists of a real-world integration

problem that builds on a project to automate the reg-

istration of new users into a unique repository of the

Huelva’s County Council. This repository contains

information about users that comes from both a local

application and a web portal. It is expected that every

new user is notiﬁed and provided with his/her digital

certiﬁcate by secure e-mail.

4.1 The Software Ecosystem

The integration solution involves six applications,

namely: Local Users, Portal Users, LDAP, Human

Resources System, Digital Certiﬁcate Platform, and

Mail Server. Each application runs on a different plat-

form, and, except for the LDAP, the Digital Certiﬁcate

Platform, and the Mail Server, they were not designed

with integration concerns in mind.

The Local Users is the ﬁrst application developed

in house; it aims to manage the county council in-

formation systems’ users. Note that, this is a stan-

dalone application and does not provide an authenti-

cation service. The Portal Users is an off-the-shelf ap-

plication that the web portal uses to manage its users.

In addition, a unique repository for users has been

set up using an LDAP-based application, so that it

can provide authentication access control for several

other applications inside the software ecosystem. The

Human Resources System is a legacy system devel-

oped in house to provide personal information about

the employees. It is a part of the integration solution

since we require information like name and e-mail

to compose notiﬁcation e-mails. Another application

developed in house is the Digital Certiﬁcate Platform,

which aims to manage digital certiﬁcates; it was de-

signed with integration concerns in mind. Amongst

other services, this application can be queried to get

a URL that temporarily points to a digital certiﬁcate

that users can download after authenticating. Finally,

the Mail Server runs the Council’s e-mail service,

which is used exclusively for notiﬁcation purposes.

4.2 Conceptual Model

The integration solution we have devised using

Guaraná is composed of one orchestration process

that exogenously co-ordinates the applications in-

volved in the integration solution. Figure 11 shows

the conceptual model of this integration solution.

Some ports use text ﬁles to communicate with Lo-

cal Users, Portal Users, and LDAP; the Human Re-

sources System is queried by means of its database

ICEIS2015-17thInternationalConferenceonEnterpriseInformationSystems

S18

Local Users

Portal Users

LDAP

Mail Server

Human

Resources

System

Digital

Certificate

Platform

T12

S10

S11

S12

S13

S14

S15

S16 S17

S19

S20

S21

T10

T11

T13

S22

Figure 11: The Huelva’s County Council conceptual model for the integration problem.

management system; and, the communication with

the Digital Certiﬁcate Platform and the Mail Server is

performed by means of APIs. Translator tasks were

used to translate messages from canonical schemas

into the schemas with which the integrated applica-

tions work.

The workﬂow begins at entry ports P1 and P2,

which periodically poll the Local Users and Portal

Users logs to ﬁnd new users. Inbound messages

wrapping the data that has been polled are written by

these ports to slots S1 and S2. Inside the process,

task T1 gets messages coming from both ports and

adds them to slot S3. Task T1 is a merger task, which

does not change the message content. The next task in

the ﬂow is a replicator that creates two copies of every

message it gets from slot S3, so that one copy can be

used to query application Human Resources System

by means of solicitor port P3 for information about

the employee who owns a user record, and the other

copy (called base message) is written to slot S7 so lat-

ter it can be correlated by task T4 with the response

from the Human Resources System. Task T3 is a

translator that change the schema of messages to cre-

ate the query message. Next, task T5 enriches the base

message with the information in the response message

and then task T7 replicates this enriched message with

copies to the LDAP and the Digital Certiﬁcate Platform.

The new user record is written to the LDAP by means

of exit port P5. Before querying the Digital Certiﬁcate

Platform, task T6 ﬁlters out messages that do not in-

clude an e-mail address. Messages that go through

task T12, which enriches them with the correspond-

ing certiﬁcate. Finally, exit port P6 communicates

with the Mail Server application to send the certiﬁ-

cate and notify the employee about his/her inclusion

in the LDAP.

4.3 Formal Model

Figure 12 shows the stochastic Petri net that models

the case study presented in the previous section. The

net was built following the translation from Guaraná

tasks to the respective Petri nets, as presented in Sec-

tion 3. These nets were connected to each other in

the same fashion of the tasks interconnections shown

in Figure 11. In the resulting Petri net, transitions

were named according to their corresponding tasks,

and places were named according to their respective

slots.

All information needed to create the stochastic

Petri net was available in the original Guaraná con-

ceptual model, with the exception of the ﬁring rates

of the transitions. In this work, in order to evaluate

the impact of the message processing across the net-

UsingPetriNetstoEnabletheSimulationofApplicationIntegrationSolutionsConceptualModels

Figure 12: Corresponding Stochastic Petri net for the integration problem.

work, all ﬁring probabilities were set to 1 with the ex-

ception of the ﬁlter task, represented by the transitions

T6a and T6b, that were set to 1.0 and 0.1 respectively,

simulating a ﬁltering of 10% of messages.

5 ANALYSIS

In order to analyse the behaviour of the Petri net

system, simulations were done using the PIPE soft-

ware (Pere et al., 2007). Twenty independent simu-

lations were carried out and the mean results are pre-

sented in this section. Table 1 presents the total num-

ber of tokens remaining in the Petri net places after

20,000 transition ﬁrings. The tokens in the Petri net

are analogous to the messages in the modelled inte-

gration solution, and the places are analogous to slots,

that act as buffers, retaining unprocessed messages.

We can see that place S3 retained a signiﬁcant num-

ber of tokens. This place models the slot right after

the merger task that receives input messages in paral-

lel. This result is consistent to the notion that a critic

point in a system is processing of incoming messages.

In the simulated system, the sum of the probability ra-

tios of the parallel input transitions for the merger task

represented by transitions T1a and T1b, is two times

bigger then the probability of the transition T2, right

after place S3, as can be seen in Table 2.

Table 2 shows the transitions and their respective

ﬁrings. Different from all other transitions, transi-

tion T6b has a global probability of 0.00497, which

Table 1: Number of tokens in places after 20,000 ﬁrings.

Place Average total tokens

S1 63.36184

S2 41.46832

S3 1005.9868

S4 22.0765

S5 64.1405

S6 22.48608

S7 108.70306

S8 28.79736

S9 28.79736

S10 27.72302

S11 21.73742

S12 28.34858

S13 6.96606

S14 9.29324

S15 9.8257

S16 6.96136

S17 8.99286

S18 25.77992

S19 6.77466

S20 6.77466

S21 7.33414

S22 18.0256

corresponds to ﬁltering 10% of messages. The table

presents the total number of ﬁrings for each transition

along with the expected ﬁrings number, according to

the transition global probability. The Firing rate col-

umn is the rate of the Total ﬁrings in relation to Ex-

pected ﬁrings column. We can see that the transitions

with the higher ratios are those located at the begin-

ICEIS2015-17thInternationalConferenceonEnterpriseInformationSystems

Table 2: Number of tokens in places after 20,000 ﬁrings.

Transition Firings Expected ﬁrings Firing rate Global probability

T1a 1059.09 995.024 106.44% 0.0497

T1b 1114.86 995.024 112.04% 0.0497

T2 1167.97 995.024 117.38% 0.0497

T3 1145.89 995.024 115.16% 0.0497

T4 1059.26 995.024 106.46% 0.0497

T5 1030.47 995.024 103.56% 0.0497

T6a 911.06 995.024 91.56% 0.0497

T6b 84.71 99.500 85.14% 0.00497

T7 1002.74 995.024 100.78% 0.0497

T8 981.01 995.024 98.59% 0.0497

T9 901.77 995.024 90.63% 0.0497

T10 891.95 995.024 89.64% 0.0497

T11 875.99 995.024 88.04% 0.0497

T12 869.22 995.024 87.36% 0.0497

T13 861.88 995.024 86.62% 0.0497

P1 1178.22 995.024 118.41% 0.0497

P2 1100.56 995.024 110.61% 0.0497

P3 1081.75 995.024 108.72% 0.0497

P4 884.98 995.024 88.94% 0.0497

P5 952.66 995.024 95.74% 0.0497

P6 843.86 995.024 84.81% 0.0497

Total 20,000 20,000 – –

ning of the chain of events, while the transitions with

lower ratios are those at the end of the chain of events.

6 CONCLUSIONS

Guaraná technology offers a set of tools and method-

ologies to support the design and implementation

of enterprise application integration solutions. Cur-

rently, the simulation approach adopted by software

engineers to analyse an integration solution depends

on the activities related with the construction, execu-

tion, and collection of data from such execution. It is

recognised that there is a high cost, risk and develop-

ment time spent associated with these activities. The

simulation approach proposed in this paper aims at

coping with this problem, by offeringa way to analyse

integration solution characteristics in an early phase

of the development cycle.

In this paper, we proposed a simulation approach

using Petri nets, in order to be able to analyse Guaraná

solutions based only on the conceptual models. Petri

nets are frequently used to model discrete event sys-

tems of any kind, which is the case of integration so-

lutions. This work shows that Petri nets are an use-

ful tool to simulate integration solutions in software

engineering. It was demonstrated that the Guaraná

conceptual model can be translated into a stochastic

Petri net and simulated, enabling to identify its char-

acteristics. Important aspects such as message staking

in buffer slots, concurrency, tasks over-demand and

number of processed messages can be observed using

Petri net simulation tools, such as PIPE. The simula-

tion of integration solutions reduces the costs by mak-

ing possible to identify possible design ﬂaws ahead of

software design and prototype.

The correct identiﬁcation of the problem leads to

precise deﬁnition of the objectives, and the type of

simulation model that must be developed. In our ap-

proach, we used simulation to understand the possi-

ble problems related to integration solutions. We con-

sider that the simulation approach presented in this

paper can improve the quality of enterprise applica-

tion integration solutions developed using Guaraná.

We noticed that the area of Enterprise Application

Integration in combination with the area of simula-

tion is poorly investigated. Therefore, this work al-

lowed the opening of promising research lines within

the scope of application integration solutions using

discrete-event simulation.

ACKNOWLEDGEMENT

The research work on which we report in this paper

is supported by CAPES, FAPERGS, and the internal

Research Programme at UNIJUI University.

UsingPetriNetstoEnabletheSimulationofApplicationIntegrationSolutionsConceptualModels

REFERENCES

Alla, H. and Ghomri, L. (2012). Modeling and Simulation

by Hybrid Petri Nets. In Proceedings of the Winter

Simulation Conference (WSC), pages 1–8.

Bause, F. and Kritzinger, P. S. (1998). Stochastic Petri Nets:

An Introduction to the Theory. SIGMETRICS Per-

form. Eval. Rev., 26(2):2–3.

Cavalieri, S. (2000). Improving Performance of a Flexi-

ble Manufacturing System by Petri Net Based Mod-

elling and Simulation. In Industrial Electronics Soci-

ety, 2000. IECON 2000. 26th Annual Confjerence of

the IEEE, volume 2, pages 1298–1303 vol.2. IEEE.

David, R. and Alla, H. (1994). Petri Nets for Modeling of

Dynamic Systems: a Survey. Automatica, 30(2):175–

202.

Davidrajuh, R. (2008). Developing a New Petri Net Tool

for Simulation of Discrete Event Systems. In Model-

ing Simulation, 2008. AICMS 08. Second Asia Inter-

national Conference on, pages 861–866. IEEE.

Dossot, D. and D’Emic, J. (2009). Mule in Action. Man-

ning.

Fisher, M., Partner, J., Bogoevici, M., and Fuld, I. (2010).

Spring Integration in Action. Manning.

Frantz, R. Z. and Corchuelo, R. (2012). A Software De-

velopment Kit to Implement Integration Solutions. In

27th Symposium On Applied Computing, pages 1647–

1652.

Frantz, R. Z., Reina-Quintero, A. M., and Corchuelo, R.

(2011). A Domain-Speciﬁc Language to Design En-

terprise Application Integration Solutions. Interna-

tional Journal of Cooperative Information Systems,

20(2):143–176.

Gold, R. (2004). Petri Nets in Software Engineering. Work-

ing Papers.

Hohpe, G. and Woolf, B. (2003). Enterprise Integration

Patterns - Designing, Building, and Deploying Mes-

saging Solutions. Addison-Wesley.

Ibsen, C. and Anstey, J. (2010). Camel in Action. Manning.

Kuo, C.-H. (2004). Development of Distributed Agent-

oriented Petri Net Simulation and Control Environ-

ment for Discrete Event Dynamic Systems. In Sys-

tems, Man and Cybernetics, 2004 IEEE International

Conference on, volume 5, pages 5001–5006 vol.5.

IEEE.

Marsan, M. A. (1990). Stochastic petri nets: An elemen-

tary introduction. In Rozenberg, G., editor, Advances

in Petri Nets 1989, pages 1–29. Springer-Verlag New

York, Inc., New York, NY, USA.

Mens, T. and Gorp, P. V. (2006). A Taxonomy of Model

Transformation. Electron. Notes Theor. Comput. Sci.,

152:125–142.

Molloy, M. K. (1981). On the Integration of Delay and

Throughput Measures in Distributed Processing Mod-

els. PhD thesis, University of California, Los Angeles.

Narahari, Y., Suryanarayanan, K., and Reddy, N. (1989).

Discrete Event Simulation of Distributed Systems Us-

ing Stochastic Petri Nets. In TENCON ’89. Fourth

IEEE Region 10 International Conference, pages 622–

625. IEEE.

Pere, B., Lladó, C. M., Puijaner, R., and Knottenbelt, W. J.

(2007). PIPE v2. 5: A Petri Net Tool for Performance

Modelling. In Proc. 23rd Latin American Conference

on Informatics (CLEI 2007).

Rezai, M., Ito, M., and Lawrence, P. (1995). Modeling and

simulation of hybrid control systems by Global Petri

Nets. In Circuits and Systems, 1995. ISCAS’95., 1995

IEEE International Symposium on, volume 2, pages

908–911 vol.2. IEEE.

Russell, N., van der Aalst, W., and ter Hofstede, A. (2009).

Designing a workﬂow system using coloured petri

nets. In Jensen, K., Billington, J., and Koutny, M.,

editors, Transactions on Petri Nets and Other Models

of Concurrency III, volume 5800 of Lecture Notes in

Computer Science, pages 1–24. Springer Berlin Hei-

delberg.

Schmidt, D. C. (2006). Guest editor’s introduction: Model-

driven engineering. IEEE Computer, 39(2):25–31.

van der Aalst, W., Stahl, C., and Westergaard, M. (2013).

Strategies for modeling complex processes using col-

ored petri nets. In Jensen, K., van der Aalst, W., Balbo,

G., Koutny, M., and Wolf, K., editors, Transactions

on Petri Nets and Other Models of Concurrency VII,

volume 7480 of Lecture Notes in Computer Science,

pages 6–55. Springer Berlin Heidelberg.

Weiss, J. (2005). Aligning Relationships: Optimizing the

Value of Strategic Outsourcing. Technical report,

IBM.

Zhou, M. (1998). Modeling, Analysis, Simulation,

Scheduling, and Control of Semiconductor Manufac-

turing Systems: A Petri Net Approach. Semiconduc-

tor Manufacturing, IEEE Transactions on, 11(3):333–

357.

ICEIS2015-17thInternationalConferenceonEnterpriseInformationSystems