On the Performance Prediction Capabilities of the eBPMN-based Model-driven Method for Business Process Simulation Paolo Bocciarelli Andrea Giglio Andrea D’Ambrogio Emiliano Paglia Guglielmo Marconi University Dept. of Innovation and Information Engineering Department of Enterprise Engineering Via Plinio 44 University of Rome Tor Vergata 00193, Rome, Italy Via del Politecnico 1 a.giglio@unimarconi.it 00133, Rome, Italy {paolo.bocciarelli, dambro, emiliano.paglia}@uniroma2.it Abstract—The analysis of business processes may take consid- by one or more organizations working together to achieve a erable advantage by the adoption of simulation-based techniques common business purpose [31], [25]. from the early phases of process lifecycle. Despite the fact that From an operational perspective, several modeling lan- Modeling & Simulation (M&S) approaches have been widely recognized as a valuable solution, the remarkable effort required guages have been introduced in the last years to specify for their implementation and their essential complexity may business processes. Among them, the Business Process Model limit their use in practice. In previous contributions we have and Notation (BPMN) [34], provided by OMG, is playing a proposed a model-driven method and a domain specific language, primary role as the de facto standard in the BPM domain. named eBPMN, for enabling the automated development and BPMN is widely adopted by users with different backgrounds execution of business process simulations. In this paper, we analyze the performance prediction capabilities of the current and roles, ranging from business analysts to IT experts [32]. implementation of the eBPMN-based model-driven method, by Since the complexity of modern BPs is constantly increas- means of a comparison with the same capabilities provided by ing, it becomes essential to use innovative methodologies and similar existing tools, such as BIMP and Bizagi, for a reference techniques to effectively support BP analysis activities right business process. from the first phases of the BP lifecycle. Index Terms—Business Process Model and Notation (BPMN), Business Process Management, Modeling and Simulation, Busi- In this respect, the use of Modeling and Simulation (M&S) ness Process Analysis techniques has been identified as an effective BP analysis approach [43], providing a valid support at both design time Copyright c held by the author and operations time. Although M&S approaches are widely recognized as ex- I. INTRODUCTION tremely valuable [20], their practical use is not yet widespread Modern organizations have to face a constantly evolving as one would expect, due to two main factors. On the one market, characterized by rapidly changing demands and in- hand, the cost to set up and to maintain a M&S environment creasing levels of competition. In such a context, it becomes is often not-negligible. On the other hand, the use of M&S- crucial for organizations to get a deep knowledge of their based techniques requires skills and know-how which most operational business model and thus have the ability to dy- business analysts are not familiar with. namically improve it, in order to gain a competitive advantage In previous contributions, we have introduced a framework and better exploit market opportunities. Specifically, a deep that largely reduces the effort and the cost for carrying out understanding of the operational business processes is essential simulation-based BP analysis activity. The proposed frame- to enhance organization efficiency and improve the quality of work exploits standards and technologies introduced in the delivered products and/or provided services. model-driven development field to ease and automate the As a consequence, the vast majority of organizations have simulation-based BP analysis [6]. The framework introduces embraced Business Process Management (BPM) approaches, a domain-specific language, named eBPMN, for the specifica- which include methods, techniques, and tools to support the tion and execution of BP simulation models. design, analysis, enactment and improvement of operational eBPMN allows one to carry out BP analysis activities by business processes [42]. A business process (BP) can be addressing both performance (i.e., efficiency) and reliability defined as a set of interrelated activities that are executed (i.e., failure-free behavior) properties. This paper specifically focuses on the performance prediction capabilities of the and suitability for communication), simulation capabilities current implementation of eBPMN. The paper discusses the (performance dimensions, distributions, animation, scenarios) effectiveness and validity of such predictions by means of a and output analysis capabilities (statistics, format, what-if comparison between the key performance indicators and the analysis, decision-making support). results provided by eBPMN and those provided by other well- An updated review on the role of simulation techniques know and widely used tools, such as BIMP and Bizagi, for a can be found in [28], in which authors report that discrete- reference business process. event simulation is by far the most investigated technique The comparison allows us to conclude that eBPMN enables (about 40% of the literature contributions), followed by system business analysts to get the same performance predictions of dynamics (about 15%), hybrid techniques and Monte-Carlo commonly available BP simulation tools, with the additional simulation, among the others. advantages of: An additional comparison of BP simulation tools is de- • not requiring the manual specification and/or implementa- scribed in [27], in which authors classify the reviewed tools tion of the simulation model, which can be automatically (e.g., JBoss jBPM simulation tool, Oracle Business Pro- generated from BP models specified by use of BPMN or cess Analysis Suite, SigmaFlow modeler, Arena simulation other BP modeling languages; modeling tool) in terms of licensing options and addressed • providing both sequential and distributed simulation exe- application domains. cution engines, which can be easily integrated into MSaaS Beyond the various features and capabilities of the several (M&S as a Service) platforms. BP simulation tools reviewed in the aforementioned contri- The rest of this paper is organized as follows: Section II pro- butions, what is considered as an actual impediment to the vides the necessary background and discusses related works. effective use of M&S-based approaches in the BPM domain Section III outiles the aforementioned features of eBPMN is the semantic gap that often exists between the BP-based and illustrates the comparison approach, while Section IV conceptual model and the BP simulation model. introduces the case study that has been used as a reference From a practical perspective, in [16] the authors observe scenario for BP performance analysis. Section V discusses the that a main focus of research is to provide simulation building related results and, finally, Section VI gives final remarks. blocks that are as close as possible to the elements of the conceptual model. The authors point out that a slight difference II. BACKGROUND between the conceptual model and the simulation language The reasons for adopting simulation techniques in the BPM can lead to difficulties in translating the model and, in some domain have been investigated in [38], [37]. Typically, simula- cases, to the impossibility to carry out such translation. In this tion techniques can answer questions such as [23], [26]: what respect, in [14] the author argues that most of the currently are the total process time and the maximum throughput of the available simulation tools provide hard-coded simulation ca- process, what are the equipment and technology requirements pabilities, failing to explicitly define the underlying simulation to meet the service demand, what are the waiting times, what modeling formalism by use of, e.g., a well defined metamodel. is the best assignment of resources to task and how to deal Attempts to bridge the gap between conceptual and simula- with unexpected situations or emergencies. tion models are documented in [40], [30], [14], [39]. Such at- However, the potential of M&S approaches is not fully tempts are based either on the provision of specific simulation exploited yet, mainly due to the fact that the several exist- languages, such as the event-driven Simulation Activity Dia- ing simulation tools require conceptual and technical skills gram (SAD) language introduced in [40], or on more flexible beyond those generally available in business organizations and effective approaches based on model-driven development [2]. In this regard, a survey conducted among potential BP and automated model transformations, such as in [30], which simulation users reveals that almost 80% of respondents do not proposes a model transformation approach that can be used to use simulation techniques, while actual users report as main translate conceptual process models, expressed in the event- motivations the ability of simulation-based analysis to support driven process chain (EPC) notation, to different simulation extensive experimentation and get a deeper understanding of software systems, and in [14], which presents a framework complex process interactions [33]. that exploits metamodeling and model transformations in order A systematic review of business process simulation tools to translate conceptual elements into executable simulation can be found in [29], in which the authors evaluate their components. Similarly, in [39] a model-driven approach is applicability, underline some limitations and derive recom- used to transform BPMN models into DEVS models and mendations for further research. The review analyzes business then to Java classes used for the simulation with the DSOL process modeling tools that may be applicable for simulation (Distributed Simulation Object Library) library. (e.g., Protos, ARIS), tools that expose simulation capabilities To summarize, the analysis of the literature reveals a con- (e.g., FLOWer, FileNet) and general purpose simulation tools siderably wide offer of M&S tools for BP simulation, but still (e.g., Arena, Colored Petri-Net tools). The evaluation is based a limited use of the potential offered by M&S approaches for on modeling capabilities (ease of model building, formal BP analysis. Further, several existing simulation tools require semantics and verification of correctness, workflow patterns, specific know-how in M&S, as well as software engineering resource and data perspective, level of detail, transparency skills, far beyond those generally available in business organi- zations. As such, model-driven development can be considered • number of tokens processed by each element and by each one of the most effective approaches to provide the required resource as a measure of the throughput; degree of automation, which helps to reduce the effort and • tokens managed on each branch of gateways as a measure cost of M&S-based BP analysis. of the usage of each possible business process flow; In this respect, we have proposed a full featured model- • cycle time (mean and variance) as the time spent by a driven framework that introduces a set of automated model token to complete the collaboration from token generation transformations to translate BP models into corresponding at start node to token termination at end node. BP simulation models and simulation model implementations Although eBPMN allows to simulate both the performance ready to be executed on top of sequential or distributed sim- and the reliability behavior of the business process [6], in this ulation engines [10], [9]. The framework includes a domain- paper we specifically focus on performance properties. This specific language, named eBPMN, to specify and execute BP is due to the fact that the proposed approach for performance simulation models automatically generated from BP models prediction analysis is based on comparing the eBPMN results specified in a given BP modeling language, such as BPMN or with those obtained with other existing BPMN simulators UML. which do not provide process reliability simulation. III. THE E BPMN DOMAIN-SPECIFIC LANGUAGE B. eBPMN-based Model-Driven Method FOR BP SIMULATION One of the main advantages of eBPMN is that the executable A. eBPMN Overview code can be directly obtained in an automated fashion from eBPMN is a domain-specific simulation language based on standard BPMN models by use of a model-driven method the execution semantics defined in the BPMN 2.0 specification based on automated model transformations [10], [9]. [34]. It has been originally introduced in [12] and further As aforementioned, such transformations rely on a extended in [9], [10], [15], [1]. lightweight extension of the BPMN metamodel, named The eBPMN language has been built on top of SimArch PyBPMN, that has been defined in previous works [7] as a [21], a layered software architecture which gives users the semantics preserving extension that does not alter the original ability to specify event-driven simulation models that can be BPMN metamodel. PyBPMN gives the ability to annotate transparently executed either in local or distributed simulation standard BPMN models with performance-oriented BP prop- environments. A detailed description of SimArch is given in erties. The annotated BPMN model is then directly mapped to [22]. eBPMN executable code. eBPMN allows users to simulate BPs consisting of a single The PyBPMN metamodel introduces specific metaclasses participant as well as complex process collaborations. A BP for the following components: element (task or activity) makes use of a single resource • workload definition: responsible for modeling the work- or more resources to perform its job. In order to specify load related to the whole business process or to the tasks the resources behavior and the non-functional properties of associated to the process (i.e., the execution of single BPMN elements, eBPMN exploits a lightweight BPMN exten- activities); sion named PyBPMN (Performability-enabled BPMN), briefly • performance properties definition: responsible for speci- summarized in next sub-section and detailed in [7]. fying the performance properties associated to both the The eBPMN language implements a token abstraction to process and the single task. The most common perfor- simulate the execution of a BPMN process. A token is gener- mance properties are the service demand (service time), ated by a Start node and can be considered as a reference the time spent to accomplish the demand (response time), to the execution of a process instance. The time interval and the throughput; between two subsequent token generation events follows a • reliability properties definition: responsible for modeling given probability distribution. After creation, the token goes the reliability related properties of the resources involved throughout the process nodes, guided by the sequence flows, in a process or associated with a task. The most common and each eBPMN element handles the token according to its reliability properties are the occurrence rate of the failure, execution semantics. At the End node the token is destroyed the occurrence distribution of the failure, the mean time and the simulation environment gathers information about the to failure (MTTF) and the mean time to repair (MTTR); process traversal. • resource management: responsible for specifying the The eBPMN simulation language provides the following actual resource which is used to execute an activity. performance metrics for process elements and collaborations: PyBPMN allows to define non functional properties for • service time (mean and variance) and waiting time (mean atomic resources, as well as groups of resources consist- and variance) for resources and tasks; ing of concurrent or alternative resources. • resource utilization, i.e., the time spent by a resource The eBPMN-based model-driven method has been spec- in executing service requests, considering the amount ified using languages and tools introduced by the OMG’s of parallel working units and thus evaluating bottleneck MDA (Model Driven Architecture) incarnation of model- issues; driven development principles, and implemented within the Eclipse Modeling Framework (EMF) platform [18], [41] and IV. CASE STUDY FOR PERFORMANCE the Eclipse Modeling Project [19], [24]. The following Eclipse VALIDATION plugins have been implemented in order to provide a tool-chain The process considered for eBPMN validation is a health that eases the production of the eBPMN-based simulation code care process, specifically a process dealing with diabetes from a standard BPMN model: care. Diabetes of type 1 must be treated with medicines • a plugin implementing the PyBPMN metamodel; throughout life. Affected patients should take insulin to avoid • a plugin implementing the BPMN to PyBPMN model-to- excessive glycemic peaks. Pharmacological treatment aims at model transformation; controlling the symptoms of diabetes and preventing serious • a plugin implementing the PyBPMN to eBPMN model- complications. to-text transformation. The case study refers to the request of essential products The tool-chain is being ported and deployed into cloud- for diabetes care. The start event consists of the expiration based environments, according to the MSaaS paradigm. The of the periodic treatment of a patient, which results in the proposed MSaaS platforms includes the aforementioned model need to contact the salesman to acquire new medicines. The transformation services, a SimArch-based simulation execution salesman, following the patient necessary quantities, prepares engine, which allows users to execute eBPMN simulations on a new medicines plan, which has then to be approved by top of either sequential or distributed simulation environments, the reference doctor (i.e., the diabetologist). After the plan and modeling services to directly specify eBPMN simulation approval, the salesman forwards the order to the pharmaceuti- models, as described in [11], [8], [5]. cal industry that packages the medicines. The pharmaceutical industry then sends the medicines to the main hospital, which C. eBPMN Performance Prediction locates the authorized medical center closest to the patient In order to analyze the eBPMN performance prediction facility and forwards the medicines to it. Finally, the medical capabilities, an indirect approach has been adopted. The center notifies the patient of the availability of the requested analysis is carried out by comparing the results provided by medicines and the process ends. Both the medicines plan and eBPMN with the outputs provided, for the same process and the order can be subjected to refusal (with a 15% probability), input parameters, by two well-known tools that offer similar which leads to rework of the request. features, i.e., Bizagi Process Modeler [4] and BIMP [3]. The BPMN model of the reference process is depicted in The Bizagi Process Modeler (Bizagi in short) is a business Figure 1. process modeling and documentation tool, compliant with Each pool in the BPMN model defines a resource role the BPMN 2.0 specification, that allows to visualize, model involved in the process. Each role is associated with one and document business processes using BPMN. The Bizagi or more actual resources (performers) which are able to Process Modeler also provides simulation parameterization execute the activity. Table I summarizes the resources roles capabilities that conform to the BPSim specification [44]. and the corresponding amount of performers considered for BIMP is a simulator of business process models, free for the reference process. academic use, available either through a web-based interface or according to a simulation as a service paradigm. TABLE I R ESOURCES ( WITH QUANTITIES ) AVAILABLE FOR THE REFERENCE The comparison is carried out with respect to some specific PROCESS . performance measures, or Key Performance Indicators (KPIs), which are able to capture the essential aspects of the business Resource Quantity process performance behavior. Salesman 1 In this respect, a significant effort has been spent in order Doctor 1 Pharmaceutical Industry 3 to deal with the definition of appropriate KPIs for BPM Hospital 2 [13], [35], [36]. A typical performance metric of interest for Medical Center 1 business processes is the cycle time, i.e., the time taken to handle one token (one process instance), addressed in terms of For each activity in the process, Table II reports both the maximum value, average value or its variation over instances. average service time and the resource performing the activity. Further, cycle time can be analyzed through its constituent measures such as the service time (the time spent to actually V. RESULTS handling the token) or the waiting time (the time spent in idle The performance validation has been carried out simulating mode, either in queue or waiting for synchronization) [17]. the reference process with respect to two different scenarios For the comparison of results provided by eBPMN, BIMP A and B, which differ for the expected workload in terms and Bizagi, the following KPIs have been taken into consid- of patients requests average inter-arrival time: 4 hours for eration: scenario A and 3 hours for scenario B. Thus, scenario B deals • process cycle time; with a more intensive workload. • resource utilization; The simulation has been set-up considering exponential • activity waiting time. probability distributions for inter-arrival and activity times. Fig. 1. Reference BP model for performance prediction capabilities analysis. TABLE II AVERAGE ACTIVITY EXECUTION TIMES AND RESOURCES IN CHARGE . Activity Avg. time [min] Resource Gather therapeutic plan 15 Salesman Check therapeutic plan 45 Doctor Approve plan 20 Doctor Process order 60 Salesman Check order 5 Pharmaceutical Industry Prepare medicine 360 Pharmaceutical Industry Send medicines to hospital 60 Pharmaceutical Industry Send medicines to medical center 180 Hospital Notify patient of medicines availability 60 Medical Center As for the simulation duration, both eBPMN and Bizagi reasonable trade-off between minimizing the statistical error allow to stop the simulation after a given simulated time, while and limiting the simulation complexity (eBPMN and Bizagi BIMP simulations are stopped when all the tokens generated provide results in about 2 minutes on a typical desktop PC by start nodes are terminated at the end nodes. The use of with 4th-generation Intel i5 CPU and 8 GByte of RAM). probability distributions for inter-arrival time, activities dura- Moreover, it should be noted that eBPMN and Bizagi tion and gateways routing policies introduces some uncertainty provide repeatable deterministic simulations while BIMP does on precision of the provided results. Indeed, if the termination not, so repeating the BIMP simulation could lead to different condition is the maximum simulated time, the exact number values. of processed tokens is not exactly predictable; on the other The KPIs provided by eBPMN, BIMP and Bizagi for end, if the termination condition is the number of tokens, the scenario A are depicted in Figure 2, while Figure 3 presents the actual simulated time is not exactly predictable. same KPIs for scenario B. Corresponding figures use the same However, if the simulated time is long enough, all the results scale to make easier the comparison of the KPI on workload tend towards the steady-state conditions of the process. This variation. allows to compare the results of different tools even if they The simulation of the two scenarios shows comparable have simulated a different amount of process instances. performance predictions of process times, resources utilization For the purpose of this paper, the simulation stops after 365 and waiting times for eBPMN, BIMP and Bizagi. Differences simulated days for eBPMN and Bizagi (about 2000 tokens in some values are likely due to pseudo-random number processed per run) and after 3000 tokens for BIMP (to meet generation for probability distribution approximation. tool constraints on simulation duration). Further, eBPMN and As expected, increasing the workload (i.e., switching from Bizagi automatically execute 50 runs of the process to reduce scenario A to scenario B) leads to higher resource utiliza- statistical error. BIMP does not allow to specify the number tion. The pharmaceutical industry has the highest utilization of runs so the simulation has been manually re-run to evaluate (from 60% for scenario A to 80% for scenario B). Under average values. these conditions, the resource is working near its maximum The aforementioned quantities have been chosen as a capacity and the waiting times for activities performed by such eBPMN BIMP Bizagi 0 2 4 6 8 10 12 14 16 18 20 22 24 26 28 30 32 34 36 38 40 Average cycle time [h] 100% eBPMN BIMP Bizagi 80% 60% 40% 20% 0% Salesman Doctor Pharmaceutical Hospital Medical Industry Center Resources utilization 4 eBPMN BIMP Bizagi 3 2 1 0 Approve Process Prepare Gather Check Check Send Send Notify therapeutic therapeutic plan order order medicine medicines medicines patient of plan plan to hospital to medical medicines center availability Activity waiting time [h] Fig. 2. Comparison of KPIs for reference process considering patient requests inter-arrival time of 4 hours. resource quickly go up. This behavior is captured, with minor provided by similar tools, i.e., BIMP and Bizagi Process differences, by all the considered tools (for the Check order Modeler, for a reference BP. The validation procedure has activity eBPMN shows higher waiting times than the other been applied to two scenarios for the reference process with two tools, while for Prepare medicines and Send medicines to different workload characterization. The analysis of the re- hospital Bizagi predicts lower waiting times than eBPMN and sults reveals comparable performance predictions for eBPMN, BIMP). BIMP and Bizagi for cycle time, resource utilization and waiting times. VI. CONCLUSIONS The paper presents an analysis of the performance predic- The comparison has showed that eBPMN enables business tion capabilities for the current implementation of eBPMN, analysts to get the same performance predictions of similar a domain-specific simulation language for BP simulation. BP simulation tools, with the additional advantages of not The main advantages of eBPMN is its formal specification, requiring the manual specification and/or implementation of compliant with the execution semantic of BPMN, and the the simulation model, which can be automatically generated ability to automatically generate the eBPMN executable code from BP models specified by use of BPMN or other BP mod- applying a model-driven tool-chain to the BPMN model. eling languages, and using a sequential/distributed simulation The analysis of the BP performance prediction has been execution engine, which can be easily integrated into MSaaS carried out by comparing the results of eBPMN with those (M&S as a Service) platforms. eBPMN BIMP Bizagi 0 2 4 6 8 10 12 14 16 18 20 22 24 26 28 30 32 34 36 38 40 Average cycle time [h] 100% eBPMN BIMP Bizagi 80% 60% 40% 20% 0% Salesman Doctor Pharmaceutical Hospital Medical Industry Center Resources utilization 4 eBPMN BIMP Bizagi 3 2 1 0 Approve Process Prepare Gather Check Check Send Send Notify therapeutic therapeutic plan order order medicine medicines medicines patient of plan plan to hospital to medical medicines center availability Activity waiting time [h] Fig. 3. Comparison of KPIs for reference process considering patient requests inter-arrival time of 3 hours. R EFERENCES Business Processes. In Proceedings of the 2014 Winter Simulation Conference, pages 3012–3023, Savannah, Georgia, 2014. IEEE Press. [1] G. Antonacci, A. Calabrese, Andrea D’Ambrogio, A. Giglio, B. Intrigila, [7] Paolo Bocciarelli, Andrea D’Ambrogio, Andrea Giglio, and Emiliano and N.L. Ghiron. A bpmn-based automated approach for the analysis of Paglia. A BPMN Extension to Enable the Explicit Modeling of Task healthcare processes. In Gaaloul W. Reddy S.M., editor, Proceedings - Resources. In Proceedings of the INCOSE Italy Conference on Systems 25th IEEE International Conference on Enabling Technologies: Infras- Engineering, Turin, Italy, 2016. tructure for Collaborative Enterprises, WETICE 2016, pages 124–129. [8] Paolo Bocciarelli, Andrea D’Ambrogio, Andrea Giglio, and Emiliano Institute of Electrical and Electronics Engineers Inc., 2016. Paglia. Model Transformation Services for MSAAS Platforms. In [2] K. D. Barber, F. W. Dewhurst, R. L. D. H. Burns, and J. B. B. Proceedings of the 2018 Spring Simulation Conference (SpringSim’18), Rogers. Business-process modelling and simulation for manufacturing Baltimore, MD, USA, 2018. management: A practical way forward. Business Process Management [9] Paolo Bocciarelli, Andrea D’Ambrogio, Andrea Giglio, Emiliano Paglia, Journal, 9(4):527–542, 2003. and Daniele Gianni. A Transformation Approach to Enact the Design- [3] BIMP. The Business Process Simulator, oct 2016. Time Simulation of BPMN Models. In 2014 IEEE 23rd International [4] Bizagi. Bizagi Process Modeler, oct 2016. WETICE Conference, pages 199–204. IEEE, 2014. [5] Paolo Bocciarelli, Andrea D’Ambrogio, Andrea Giglio, and Antonio [10] Paolo Bocciarelli, Andrea D’Ambrogio, Andrea Giglio, Emiliano Paglia, Mastromattei. Automated Development of Web-based Modeling Ser- and Daniele Gianni. Empowering Business Process Simulation Through vices for MSaaS Platforms. In International Workshop on Model- Automated Model Transformations. In Proceedings of the Symposium Driven Approaches for Simulation Engineering (Mod4Sim 2017), part of on Theory of Modeling & Simulation - DEVS Integrative, DEVS ’14, the 2017 Spring Simulation Multiconference (SpringSim 2017), Virginia pages 39:1–39:9, Tampa, Florida, 2014. Beach, VA, USA, 2017. [11] Paolo Bocciarelli, Andrea D’Ambrogio, Andrea Giglio, Emiliano Paglia, [6] Paolo Bocciarelli, Andrea D’Ambrogio, Andrea Giglio, and Emiliano and Tommaso Panetti. A Cloud-based Service-oriented Architecture for Paglia. Simulation-Based Performance and Reliability Analysis of Business Process Modeling and Simulation. In Proceedings of the 3rd INCOSE Italy Conference on Systems Engineering, pages 87–93, Naples, Conference on Business Computing and Global Informatization, pages Italy, 2017. 151–154, 2012. [12] Paolo Bocciarelli, Andrea D’Ambrogio, and Emiliano Paglia. A [36] David Parmenter. Key performance indicators: developing, implement- Language for Enabling Model-Driven Analysis of Business Processes. ing, and using winning kpis. John Wiley & Sons, 2007. In Proceedings of the 2nd International Conference on Model-Driven [37] Ray J. Paul, George M. Giaglis, and Vlatka Hlupic. Simulation of Engineering and Software Development, MODELSWARD’14, Lisbon, Business Processes. American Behavioral Scientist, 42(10):1551–1576, Portugal, 2014. SciTePress. 1999. [13] Evellin Cristine Souza Cardoso. Challenges in Performance Analysis in [38] Ray J. Paul, Vlatka Hlupic, and George M. Giaglis. Simulation modeling Enterprise Architectures. 17th IEEE International Enterprise Distributed of business processes. In Proceedings of the 3rd UK Academy of Object Computing Conference Workshops, pages 327–336, 2013. Information Systems Conference, pages 311–320, 1998. [14] Deniz Cetinkaya, Alexander Verbraeck, and Mamadou Seck. Applying [39] Igor Rust, Deniz Cetinkaya, Mamadou Seck, and Ivo Wenzler. Business a Model Driven Approach To Component Based Modeling and Simula- Process Simulation for Management Consultants: a Devs-Based Simpli- tion. In Proceedings of the 2010 Winter Simulation Conference, pages fied Business Process Modelling Library. In Proceedings of the 23rd 546–553, 2010. European Modelling and Simulation Symposium (EMSS 2011), 2011. [15] Andrea D’Ambrogio, Emiliano Paglia, Paolo Bocciarelli, and Andrea [40] John Ryan and Cathal Heavey. Process modeling for simulation. In Giglio. Towards performance-oriented perfective evolution of BPMN Computers in Industry, volume 57, pages 437–450. Elsevier Science models. In 6th International Workshop on Model-Driven Approaches Publishers, 2006. for Simulation Engineering, Pasadena, CA, USA., 2016. [41] Dave Steinberg, Frank Budinsky, Marcelo Paternostro, and Ed Merks. [16] Gert Jan De Vreede, Alexander Verbraeck, and Daniel T. T. Van Eijck. EMF Eclipse Modeling Framework. 2008. Integrating the conceptualization and simulation of business processes: [42] Arthur ter Hofstede, Wil van der Aalst, and Mathias Weske. Business A modeling method and arena template. Simulation, 79(1):43–55, 2003. Process Management: A Survey. In Mathias Weske, editor, Business [17] Marlon Dumas, Marcello La Rosa, Jan Mendling, and Hajo A. Reijers. Process Management, volume 2678 of Lecture Notes in Computer Fundamentals of Business Process Management. Springer-Verlag, 2013. Science, page 1019. Springer Berlin / Heidelberg, 2003. [18] Eclipse Foundation. Eclipse Modeling Framework (EMF), sep 2016. [43] Wil M. P. van der Aalst. Business Process Management: A Comprehen- [19] Eclipse Foundation. Eclipse Modeling Project, sep 2016. sive Survey. ISRN Software Engineering, 2013:1–37, 2013. [44] WfMC. Business Process Simulation Specification. 2016. [20] A Falcone, A Garro, and A Tundis. Modeling and simulation for the performance evaluation of the on-board communication system of a metro train. In 13th International Conference on Modeling and Applied Simulation, MAS 2014, pages 20–29, 2014. [21] Daniele Gianni, Andrea D’Ambrogio, and Giuseppe Iazeolla. A layered architecture for the model-driven development of distributed simulators. In Proceedings of the 1st international conference on Simulation tools and techniques for communications, networks and systems & workshops, page 61. ICST, 2008. [22] Daniele Gianni, Andrea D’Ambrogio, and Giuseppe Iazeolla. A software architecture to ease the development of distributed simulation systems. Simulation, 87(9):819–836, 2011. [23] Bruce Gladwin and Kerim Tumay. Modeling Business Processes with Simulation Tools. In Proceedings of the 26th Conference on Winter Simulation, pages 114–121, 1994. [24] Richard Gronback. Eclipse Modeling Project: a Domain-Specific Lan- guage Toolkit. 2009. [25] Michael Hammer and James Champy. Reengineering the corporation: A manifesto for business revolution. Business Horizons, 36(5):90–91, 1993. [26] Geoffrey Hook. Business Process Modeling and Simulation. In Proceedings of the 2011 Winter Simulation Conference, pages 773–778, 2011. [27] Ibrahim Imam, Nihal Nounou, Alaa Hamouda, and Hebat Allah Abdul Khalek. Survey of business process simulation tools: A comparative approach. Proceedings of SPIE - The International Society for Optical Engineering, 8285(July):0–7, 2011. [28] Mohsen Jahangirian, Tillal Eldabi, Aisha Naseer, Lampros K. Ster- gioulas, and Terry Young. Simulation in manufacturing and business: A review. European Journal of Operational Research, 203(1):1–13, 2010. [29] Monique H. Jansen-Vullers and Mariska Netjes. Business Process Simulation - A Tool Survey. Workshop and Tutorial on Practical Use of Coloured Petri Nets and the CPN, 834:77–96, 2006. [30] Oliver Kloos, Volker Nissen, and Mathias Petsch. From Process to Simulation - A Transformation Model Approach. In Proceedings of the 3rd International Workshop on Enterprise Modelling and Information Systems Architectures, pages 83–96, 2009. [31] Ryan K L Ko. A computer scientist’s introductory guide to business process management (BPM). Crossroads, 15(4):4:11—-4:18, 2009. [32] Matteo Magnani and Danilo Montesi. BPDMN: A conservative exten- sion of BPMN with enhanced data representation capabilities. CoRR, abs/0907.1978, 2009. [33] Nuno Melão and Michael Pidd. Use of business process simulation: A survey of practitioners. Journal of the Operational Research Society, 54(1):2–10, 2003. [34] OMG. Business Process Modeling Notation (BPMN), version 2.0. 2011. [35] Wang Pan and He Wei. Research on Key Performance Indicator (KPI) of Business Process. In IEEE, editor, 2012 Second International