International Journal of Modeling and Optimization, Vol. 2, No.

4, August 2012

Simulating Business Processes – A Review of Tools and

Techniques
Benny Mathew and Rajesh Mansharamani

 many companies are using tools or methods for static and

Abstract—There is a great need for simulation of dynamic modeling. While static modeling is more widely
dynamic/temporal aspects of business processes. The current practiced, dynamic modeling usage is still quite low [2].
standard for modeling business processes is BPMN and in this However, it has been established that the most important
review, the various techniques and tools that use BPMN for
simulation are studied. The review also points to shortcomings
problems related to BPC projects are the inability to
in BPMN support for simulation and in the BPM tools claim to accurately predict the outcome of radical change and the
support simulation of business processes. We also present a case inability to recognize the dynamic nature of the processes.
for general purpose DES tools to widen their scope and create Simulation of business processes has been suggested for use
components/plugins to convert BPMN output to DES input. in BPC projects as it allows the essence of business systems
to be understood, the processes for change to be identified,
Index Terms—BPM, BPMN; BPD; BPC; simulation; process visions to be developed, new processes to be
business process; discrete event; operations research; DES;
DEVS. designed and prototyped and the impact of proposed changes
on key performance indicators (KPI) to be evaluated [3].
Various methods and techniques can be used for modeling
I. INTRODUCTION business processes in order to obtain an understanding of
possible scenarios for improvement. IDEF0, IDEF3, Petri
Since the early days of computer simulation beginning in
Nets, System Dynamics, Knowledge-based Techniques and
the 1950s, discrete-event simulation (DES) has been the most
Discrete-Event Simulation are some examples of business
popular amongst various modeling techniques [1]. Consider
process modeling (BPM) techniques [4][5]. To address the
simulating flow of a particular fluid in a chemical factory. If
market requiring the improvement of BPM tools with the
the flow rate is plotted against time it will be a continuous
components for dynamic modeling, BPM software tools
curve and hence not suited for DES. On the other hand, DES
vendors are incorporating simulation modeling features
is suitable for problems in which variables change in discrete
usually using Discrete-Event Simulation (DES) [6].
times. For example, if the inventory of a particular item in a
Before examining the impediments in getting to DES from
warehouse is plotted against time, it will be a stepped one. So
BPMN, let‟s examine both technologies in greater depth.
the increase or decrease of inventory is a discrete function.
Apart from inventory management, DES has been found to
be invaluable for simulation of hospital patient management,
III. OVERVIEW OF DES AND BPMN
computer systems and manufacturing plants
The Business Process Modeling Notation (BPMN) is a A. DES Overview
standard to model business process flows and web services. DES models systems as a network of queues and activities,
Initially created by the Business Process Management where state changes occur at discrete points of time. Entities
Initiative (BPMI) and now handled by Object Management represent objects or individuals and they move through the
Group (OMG), the goal of BPMN is to provide a notation that system and generally are either receiving some sort of
is readily understandable by all business users. This includes time-consuming service or are waiting to receive service [7].
the business analysts that create the initial drafts of the Graphically the structure of a discrete event simulation
processes, to the technical developers responsible for model resembles a flowchart. The nodes are linked by
implementing the technology that will perform those directed arcs that represent the flow of entities through the
processes and finally, to the business staff who will manage system. The nodes correspond to resources where entities
and monitor those processes. Thus, BPMN creates a wait, receive service or get directed to the next resource.
standardized bridge to address the gap between the business Movement of entities over the arcs can represent physical
process design and process implementation. movement in the system or a change in status like from
waiting for service to receiving service. Entities, as they
move through a simulation model, frequently need to carry
II. NEED FOR SIMULATING BUSINESS PROCESS along with them pieces of descriptive information such as a
In order to view impact of business process change (BPC), part number, a customer preference, a classification for a
package, and so on. These pieces of information, known as
Manuscript received May 12, 2012; revised June 25, 2012.
attributes, are used at various points in the simulation model
B. Mathew is with Performance Engineering Research Centre of Tata to control the routing and handling of the entities.
Consultancy Services (email: benny1.m@tcs.com) Let‟s take an example of a supermarket. A customer enters
R. Mansharamani is currently a freelance consultant in IT Systems
Performance Engineering based out of Mumbai This work was done when
the supermarket either directly arriving at the entrance or at
he was with Performance Engineering Research Centre of Tata Consultancy. the supermarkets car-park. Within the supermarket the

417
 International Journal of Modeling and Optimization, Vol. 2, No. 4, August 2012

customer will shop for various items. All items are arranged destination based on the probability values. No queuing
on the supermarket shelves where customer directly picks happens at either the parking or the shopping area and an
them up except at the meat and the medical section. At the infinite-server model is used to emulate the time taken for
meat and medical sections the customer has to wait in queue parking and also for shopping. At the medicine and meat
for his turn. Finally the customer will pay at the checkout sections there are queues using FIFO discipline and there
counters and leave the supermarket. The queuing network may be one or more persons serving at the counters. For the
diagram that can be simulate with KPI being the end to end meat and medical section, N-server model is used. Entities
time taken by the customer from arrival at parking/entrance arriving at the checkout counters choose from one of the
to check-out is shown in Figure 1. many checkout counters, depending upon the queue size.
Here, the arrival of entities (customers in this case) is This logic is embedded in the router. The router keeps track
generated using a time based generator. There are two of the number of entities in the downstream queues and sends
directed arcs from the generator indicating that the generated subsequent entities to the smallest size queue. Each checkout
entities can either directly go to shopping area or will go to counter is represented by a FIFO discipline queue and also a
shopping area via the parking. Probabilities are assigned to single-server model. Finally the entities enter a sink where
the directed arcs the generated entities are directed to the statistics for each entity can be collected.

Parking Medicine Medical

(Infinite Queue Section 1
Server) ( FIFO ) ( N Server )

∞ N

∞ N 1

Town Shopping Area Meat Queue Meat Section Checkout

(Time-based Exit
( Infinite ( FIFO ) ( N Server ) Queue
Generator ) 1 ( Sink )
Server ) Selection
( Routing )

Checkout Checkout
Queues Counters
( FIFO ) ( Single
Servers )

Fig. 1. Example of DES Network Diagram

Car? Shop Med? Meat? Billing

Park Car Medicines Butcher

Fig. 2. Example of BPMN diagram

B. BPMN Overview supermarket will appear as given in Figure 2. This is a

BPMN defines a Business Process Diagram (BPD), which simplified example and usually a business analyst will model
is based on a flowcharting technique tailored for creating each of the activity in more details (see section VII).
graphical models of business process operations [8]. A
Business Process Model, then, is a network of graphical
IV. GAPS IN DOING DES USING BPMN
objects, which are activities (i.e., work) and the flow controls
that define their order of performance. In DES, entities move through a series of queues or buffers
A BPD is made up of a set of graphical elements which acquiring and releasing resources as they move through the
were chosen to be distinguishable from each other and to model domain [9]. The simulation is driven by a sequence of
utilize shapes that are familiar to most modelers. For example, discrete events that occur when activities are completed, and
activities are rectangles and decisions are diamonds. the entity movement that occurs as a consequence of these
Usually simulating BPMN models is concerned with events occurring. A superficial view of BPMN diagrams
verifying correctness and accuracy of the sequence flow of suggests that this is quite similar to DES. Instead of
diagrams. For this purpose a token is propagated from the transactions or entities moving between blocks, we have
Start Event of the diagram along Sequence flows, across tokens moving between symbols. However it is apparent that
Activities and Gateways, being duplicated and merged when this is not the case.
necessary, until it is consumed by an End Event. By studying Let us examine some of these differences:
the path of the token one can verify if the modeled process  Unlike in DES, in BPMN there is no equivalent of a
flow behaves in the same manner that the decision maker queue. In DES, queues are resources where the
wishes or what changes are required to obtain the desired transactions or entities wait before acquiring resources or
flow. For each activity one can associate a resource or messages.
performer (human resource).  There is no provision of assigning priorities to a token.
Taking the same supermarket example, the BPD of the  Once an activity is started there is no way another token

418
 International Journal of Modeling and Optimization, Vol. 2, No. 4, August 2012

can pre-empt the activity VI. BUSINESS PROCESS SIMULATION TOOLS

 Though each activity can be assigned a time for There are several BPM tools that support simulation and
completion, there is no way of assigning inter-task times. some of the tools are listed in Table 1. Usually BPM tools
 In DES wherever branching occurs, probabilities are they have a diagramming environment where BPDs are
assigned to each branch. However in BPMN there is no created and properties assigned to the activities.
provision to do so. The tools also have a simulation environment where
 There is no way of specifying routing policies like temporal properties are assigned. The BPM tools employee
shortest number or round-robin in BPMN various techniques in order to support simulation which are
outside the boundaries of BPMN.
Though the level of simulation support varies from tools to
V. RESEARCH WORK ON BRIDGING THE GAP tool, the more sophisticated tools like L-SIM [9] employs the
There has been extensive research in the area of making following techniques:
business process models more conducive for simulation.  Allow for probabilities to be assigned wherever token
Some of these approaches are even prior to BPMN becoming split at gateways. Though this appears simple, BPMN has
the standard for modeling business processes. different types of gateways depending on which type of
Syrjakow [10] converts a business activity and makes it gateway is used for splitting and merging appropriate
compatible with DES. Each business activity is converted mechanisms need to be used during simulation.
into event with begin and end as well as requiring resources.  Identify symbols where this is possibility of queuing and
Their web-based tool converts business process output in get inputs from user about the queuing discipline.
XML format to Stochastic Petri Nets (SPN). Temporal  Identify symbols where resource/performers are required
properties are assigned post conversion to SPN. and get time is required for the activity in the form of
distribution like normal or exponential.
Garćıa-Baňuelos et al. [11] have developed tool called
OXProS that converts BPMN generated XML output to  Some tools have built in optimizers and expect user
entered parameters to be a range of values. The optimizer
Colored Petri-Nets (CPN). It expects simulation parameters
automatically finds out the parameter values that
like arrival times and branching probabilities in with the
optimizes the KPI.
XML using BPMNs extensibility mechanisms. OXProS
During the simulation run, various statistics can be viewed
provides necessary services for CPN simulation and produces at runtime and on completion of the run, report and various
a log file that is analyzed using offline tools to extract the key statistics of the simulation is generated.
performance indicators.
Wagner et al. [12] suggests modifying DES specifications TABLE I: BUSINESS PROCESS SIMULATION TOOLS
itself to support activity. Internally the DES till will treat Whether open
Tools Vendor
business activity as a complex DES event having a start event source
and an end event. eClarus Business Process eClarus 
Modeller
Wong et al. [13] describe a relative-timed semantic model ARIS IDS Scheer 
for BPMN in the language of Communicating Sequential Infinity Process Platform Sungard 
Processes (CSP). The model is augmented by introducing the Process Modeller itp-commerce 
notion of relative time in the form of delays chosen G2 Gensym 
non-deterministically from a range. The authors adopt a IBM Websphere Lombardi IBM 
variant of two-phase functioning approach widely used in Rational System Architect IBM 
real-time systems and coordination languages like Linda. Metastorm Provision Metastorm 
Gagnéet al. [14] suggests extensions to BPMN and calls it
Cordys Business Process Cordys 
Management Suite
Time-BPMN. Time-BPMN captures the temporal ActiveVOS Active Endpoints 
perspective of business processes. Time-BPMN allows for Arena Rockwell 
Automation
the specification and depiction of temporal constraints and Intalio|BPMS Intalio 
dependencies within a BPMN diagram. With the extensions, JBoss - jBPM Suite - 
BPMN becomes expressive enough to account for a very
large set of real world business processes that are time
dependent. VII. FURTHER CONSIDERATIONS FOR BUSINESS PROCESS
The Event-driven Process Chain (EPC) is a business SIMULATION
process modeling language that captures the temporal and
While the simulation abilities of BPM tools have no doubt
logical dependencies between activities of a business been useful, there are still a few concerns:
process. It was initially described by Keller et al. [15] and  Complexity of certain business process cannot be
then formalized by different authors [16]. EPCs offer three completely captured by BPD tools. In such case where
element types: functions, events, and connectors. EPCs the KPI is highly dependent on complex routing or
extended with data, resources, time and probabilities, called business logic, there is no alternative other than use of
extended EPCs (eEPCs) are intensively used in commercial general-purpose DES tools that allow programming in
tools like ARIS and SAP R/3 [17]. a computer language.

419
 International Journal of Modeling and Optimization, Vol. 2, No. 4, August 2012

 A Business analyst will tend to model differently as support for simulation is due to the worry that these new
compared to someone with specialized simulation extensions will make BPMN too complex for business
training/experience like Operations Research (OR) analysts. So the option of BPMN extensions for simulation
analyst. In the supermarket example, most likely a will not be available at least in the near future.
business analyst will put the customer and
B. Increase Sophistication of Business Process Simulation
supermarket into swim-lanes and model the Tools
interactions between them in more detail (Figure 3).
For example at the checkout business person will Section VI described how various tools operate outside the
model interactions like „give credit card‟ and „get BPMN boundaries and provide simulation support. These
receipt‟. So desirable feature of the simulation tools can be expected to continue to add features and
environment is the ability to merge several activities sophistications to address the current limitations. The
Customer

Place Receive Place Receive Place Items Handover Receive Receive

Meat? Med.?
Meat Order Items Med. Order Medicines on Counter Credit Card Credit Card All Items
Meat

Process Handover
Request Req. Items
Supermarket
Medicine

Process Handover
Request Req. Items
Checkout

Bill all Charge Return Handover

Items Credit Card Credit Card All Items

Fig. 3. Example of diagram using BPMN with swim-lanes

into one. drawback of this approach is that the person carrying out
 Customer activities that take time but have no simulation will be required to have expertise in both BPMN
interactions with supermarket personnel/systems, like as well as with Discrete Event Simulation.
parking car and shopping will tend to be ignored by a
C. Convert from BPMN to Discrete Event Specifications
business modeler. Here it is essential for the business
modeler to also keep in mind the simulation KPI while Another option is to model and simulate complex business
drawing the BPDs. scenarios using general purpose DES tools. The building of
models for business processes from scratch is not the ideal
 Usually the business analyst will split the
situation, since there is still great value in BPMN, and the
organizations entire business processes into smaller
efforts of the business analyst will get wasted. One way to
and easier to handle segments and draw separate BPD
avoid this is to have BPMN tool‟s output converted to
for each. For example an insurance company may sell
standard DES specification like Discrete Event System
life, automobile, and medical insurance and the
Specification (DEVS). The techniques mentioned in Section
modeler will tend to draw separate BPDs for each.
s V and VI can be used for the conversion. An OR analyst can
However, the activities in these separate BPDs may
then import the model into general purpose DES tools with
share common resources/performers. If we simulate
which he/she is more familiar with and can exploit all the
each BPD separately, the response time, utilization,
features of DES tools. This approach avoids adding
wait time in queues for shared resources will be
complicated extensions to BPMN, keeping business analysts
incorrect. This means that there should be
happy and the work of OR analysts also gets simplified since
mechanisms to import separate BPDs into the DES
they start off with a partially ready simulation model.
environment and simulate them as one.
Though this approach may seem ideal, the drawback to this
approach is that BPMN diagramming tools have not adopted
a common standard for output. This means that the
VIII. FUTURE FOR DISCRETE EVENT SIMULATION
conversion tools will have to be aware of the different output
The current limitations of using BPMN for simulation can formats of the BPMN tools. Despite this drawback, there is
be overcome by one the means given below: definitely value in DES tools to have additional
A. Extensions to BPMN plugins/components that will convert BPMN output to
Discrete Event Specifications.
The next release of BPMN is going to be BPMN 2.0 and
going by the draft specifications, there is not much of an
REFERENCES
improvement its support for simulation. The primary
[1] S. Robinson, “Discrete-event simulation: from the pioneers to the
intention of BPMN was to provide a notation that is readily present, what next?,” Journal of Operational Research Society, 2005,
understandable by business analysts and resistance to add vol. 56, pp. 619-629

420
 International Journal of Modeling and Optimization, Vol. 2, No. 4, August 2012

[2] W. Currie and V. Hlupic, “Simulation Modeling: The link between B. Mathew is a member of IACSIT (80338770). He
change management panaceas,” in Proceedings of the 2000 Winter completed his B.E. in mechanical engineering from
Simulation Conference, pp. 2022-2028 University of Pane, India in 1992. He then did M.
[3] A. Greasley and S. Barlow, “Using simulation modeling for BPR: Tech in reliability engineering from Indian Institute of
Resource allocation in a police custody process”, International Journal Technology Bombay, India in 1996.
of Operations & Production Management, 18(9/10), pp. 978-988 He is working as Senior Scientist at Tata Consultancy
[4] J. Eatock, G. M. Giaglis, R. J. Paul, and A. Serrano, “The implications Services the performance engineering research centre
of information technology infrastructure capabilities for business since the year 2009. Previously he has worked with
process change success,” P. Henderson (Ed.) Systems engineering for Tata Research Development & Design Centre, Pune
business process change, London, Springer-Verlag pp. 127-137 as a part of performance engineering group and also with Veritas Software.
[5] A. F, Seila, V. Ceric, and P. Tadikamalla, Applied Simulation In Veritas too, he was part of the performance engineering group. His current
Modeling, Southbank, Australia, Thomson Learning areas of research include discrete event simulation and modelling of
[6] C. Hall and P. Harmon, “The 2005 enterprise architecture, process computer systems, call centres and business processes. He has also worked
modelling & simulation tools report,” Business Process Trends, in the areas of performance measurement, testing and tuning of file systems,
[Online]. http://www.bptrends.com databases and middleware.
[7] Ed Hughes, “Modelling Systems and Resources with SAS Simulation Mr. Mathew is also a member of IAENG(118060).
Studio: An Introduction,” SAS Global Forum 2010
[8] S. White, “Introduction to BPMN,” Object Management Group,
Business Process Model and Notation, [Onlie]. http://www.bpmn.org
[9] A. Walker, M. Clark, and L. Enstone, “L-SIM : Simulating BPMN
diagrams with a purpose built engine,” in Proceedings of the 2006 R. Mansharamani completed his B.Tech. In
Winter Simulation Conference, pp. 591-597 computer science & engineering from Indian
[10] E. Syrjakow and M. Syrjakow, “Web-based Business Process Institute of Technology-Bombay in 1988. He then
Modelling and Optimization,” IMSA 2002: Kaua'i, Hawaii, USA, pp. went on to complete his MS and PhD in computer
124-129 sciences from the University of Wisconsin,
[11] L. Garc´ıa-Ba˜nuelos and M. Dumas, “Towards an Open and Madison, in 1989 and 1993, respectively.
Extensible Business Process Simulation Engine,” CPN Workshop He joined Tata Research Development & Design
2009, [Online]. http://www.cs.au.dk/CPnets/events/workshop09/ Centre, Pune, in 1994, where he started a
[12] G. Wagner, O. Nicolae, and J Werner, “Extending discrete event performance engineering group. In 2006, he started
simulation by adding and activity concept for business process the Performance Engineering Research Centre for Tata Consultancy
modeling and simulation,” in Proceedings of the 2009 Winter Services in Mumbai, where he was a Vice President and Chief Scientist. He
Simulation Conference, pp. 2961-296. is currently a freelance consultant in IT Systems Performance Engineering
[13] P. Wong and J. Gibbons, “Relative Timed Semantics for BPMN,” based out of Mumbai. He has published 17 papers, and 2 book chapters, and
Electronic Notes in Theoretical Computer Science, Volume 229, Issue holds 15 patents. Some of his recent publications are “Recovery from
2, (July 2009), pp. 59-75 failures due to Mandelbugs in IT Systems,” IEEE PRDC 2011 (along with K.
[14] D. Gagné and A. Trudel, “Time-BPMN,” IEEE Conference on Trivedi and others), “Performance Engineering of a Trading Exchange‟s
Commerce and Enterprise Computing 2009, pp. 361-367 Risk Management System,” CMG 2010 (along with M. Nambiar), and
[15] G. Keller, M. Nüttgens, and A. W. Scheer, “Semantische “Performance Testing: Far from Steady State”, IEEE COMPSAC 2010
Prozessmodellierung,” Ereignisgesteuerter Prozeβketten (EPK), Heft (along with A. Khanapurkar and others). His current research interests are
89, Institut für Wirtschaftsinformatik, Saarbrücken, Germany (in Performance Engineering Process in IT organizations, and Virtual
German) Production Environments.
[16] J. Mendling and W. van der Aalst, “Towards EPC Semantics based on Dr. Mansharamani was a member of the ACM in 2011. He has served on
State and Context,” in Proceedings of the 5th EPC Workshop EPK Technical Committees of several conferences. Under his supervision TCS
2006, CEUR. pp.25-48 won awards for its Java profiler Jensor “Skoch-The World is Open Award!”
[17] A.-W. Scheer, ARIS : Business Process Modeling, Springer-Verlag, in 2008, and for its WAN Emulator WANemp “FOSS India Award”.
Berlin, 2nd edition, 1998

421