The adoption of Modeling & Simulation (M&S) approaches is widely recognized as a valuable solution for enacting a timely analysis of business processes (BPs). Despite their relevance, the effective introduction of such approaches in the BP lifecycle is still limited, due to the know-how and skills for building and implementing a simulation model and to the cost and effort for setting up and maintaining the execution platform. In this respect, this paper proposes a cloud-based architecture that exploits the M&S as a Service (MSaaS) paradigm and containerization technology for the flexible and dynamic composition of M&S services, so to allow business analysts to carry out an effortless and effective M&Sbased BP analysis. An example case study dealing with an e-commerce scenario is also presented in order to show the actual application of the proposed approach.
I. INTRODUCTION
Modeling and Simulation (M&S) is widely recognized as a
valuable solution to analyze the behavior of complex systems
and processes at both design time and operation time, in order
to assess their performance and evaluate design alternatives
from the early lifecycle stages. This work specifically focuses
on the use of M&S for the analysis of Business Processes
(BPs). The BP management field particularly benefits from
the adoption of M&S-based approaches, which allow business
analysts, enterprise managers and ICT specialists to get the
appropriate performance-oriented understanding of their BPs,
according to the desired or required level of abstraction, i.e.,
from the BP conceptual definition down to the technical BP
implementation and operation [
Despite the acknowledged relevance of M&S approaches, their actual introduction in the BP lifecycle is still limited, mostly due to the significant know-how, effort and cost that are needed to build, implement, deploy and execute simulation models of BPs.
In order to overcome such limitations and provide a more effective adoption of M&S approaches for BP analysis, this paper proposes a solution based on the M&S as a Service (MSaaS) paradigm, which brings the benefits of serviceoriented architectures and cloud computing into the M&S field, so to enhance interoperability, composability, reusability and reduce the effort and cost of M&S.
MSaaS platforms enable the flexible and dynamic composition of M&S services that provide on demand access to M&S solutions to be used both at definition time, when the BP is specified by use of a BP modeling language (e.g., BPMN Business Process Model and Notation), and at operation time, when the BP execution is monitored and proper decisions have to be taken in case of a performance downgrade, so to get a performance improvement.
This paper contribution further enhances standard MSaaS capabilites by exploiting model-driven development (MDD) approaches, which allow to significantly reduce the effort of developing, deploying and executing simulations by focusing on the use of formal models and automated model transformations, as illustrated in [2] and [3].
In this respect, the main contribution of this paper is the specification of a platform, denoted as BP-MSaaS, which is founded on the use of: a cloud computing infrastructure, to outsource and to dynamically configure the required hardware and software platform without the necessity of building and maintaining costly internetworked computational nodes; a model-driven development (MDD) approach, to significantly reduce the effort of building, implementing, deploying and executing BP simulation models; a microservice paradigm, a granular service oriented architecture which allows to compose the desired M&S capability by integrating loosely coupled selfcontained services that cooperate by use of a standard communication protocol [4]; a containerization method, which allows to wrap services in a self-contained runtime environment, in order to ease their cloud-based deployment and execution. Specifically, this paper adopts Docker [5], an open-source containerization infrastructure that provides native integration with various cloud service providers, such as Microsoft Azure [6] and Amazon AWS [7].
It is worth noting that the proposed architecture is flexible enough to be effectively used not only in the BP management domain, as discussed in Section III.
In order to provide an actual understanding of the proposed contribution, the paper also presents its application to an example e-commerce scenario.
The reminder of this paper is structured as follows: Section II outlines the literature review and summarizes the identified open issues that limit the effective adoption of M&S-based BP analysis. Section III illustrates the proposed BP-MSaaS platform, which is described from both an architectural and a behavioral perspective. Section IV presents an application scenario and provides additional implementation details. Finally, Section V gives concluding remarks.
OPEN ISSUES IN BUSINESS PROCESS MODELING AND
SIMULATION
BP modeling and simulation is a widely considered and analyzed research and development field. A complete and rigorous investigation on methodologies, techniques and tools can be found in [8]. An additional analysis on modeling methods such as Gantt Chart, IDEF family methods, Colored Petri-Net, Object Oriented methods, Workflow techniques is provided in [9], while an overview on BPMN, BPEL4WS, UML, Petri-Net can be found in [10], [11]. A quantitative analysis and classification of BPMN, Flow Chart, EPC and UML modeling languages, according to the Moody’s quality criterion for a good diagram [12], is provided in [13], in which the authors state that BPMN is the language with the highest score, but still far from the maximum value.
A key point emerging from the literature review is the need to adopt approaches that consider modeling as a continuous activity to be carried out throughout the business process lifecycle, thus evidencing the strict correlation with other activities and, in particular, with simulation. Such a correlation is also discussed in [14], in which the authors present the benefits that a multi-dimensional approach provides, in [15], which presents methods to obtain a model suitable for business process simulation, and in [16], in which simulation techniques are presented as complementary to modeling, since they separately consider dynamic and static aspects of a process, respectively.
Several authors reveal some limitations of current business
process modeling and simulation approaches, such as the lack
of focus on the actual IT platform that enacts the business
process model, and the semantic gap between conceptual and
simulation models [17], [18], [19], [20], [
These limitations have been addressed in literature,
and strategies to overcome them have been presented,
such as multi-dimensional approaches [14], hybrid modeling
techniques with performance analysis and process optimization
[19], and model-driven approaches that aim at reducing the gap
between conceptual and simulation models [20], [
A systematic review of business process simulation tools
can be found in [15], in which the authors analyze business
process modeling tools that also provide simulation features,
business process management tools having simulation
capabilities and general purpose simulation tools. Several simulation
tools are also compared in [
Most of the existing simulation tools require M&S
knowhow and programming skills above those generally available
in organizations [
To summarize, the literature review has identified the following open issues as limitations of currently available approaches to M&S-based BP analysis:
BP modeling is considered as a separate activity rather than being correlated to simulation throughout the BP lifecycle; a significant semantic gap exists between the BP model and the simulation model; existing simulation tools require M&S know-how and programming skills which business analysts are usually not familiar with.
In order to overcome such limitations, this paper proposes the BP-MSaaS platform, as illustrated in next section.
III.
BP-MSAAS PLATFORM FOR M&S-BASED BP
ANALYSIS
This section describes the BP-MSaaS platform, which enables the M&S-based analysis of BPs according to a MSaaS paradigm. In this respect, Section III-A illustrates the platform architecture, whereas Section III-B gives the behavioral specification.
The proposed architecture, shown in Figure 1, has been designed in order to overcome the most relevant open issues affecting existing M&S-based BP analysis approaches, as discussed in Section II. The rationale of the proposed approach is that a set of cloud-based remote services can be selected, orchestrated and executed in order to carry out the several tasks required to conduct a BP analysis: modeling services to specify abstract BP models; transformation services to generate the corresponding executable simulation models; simulation services to execute simulations and presentation services to summarize and show the related results.
The outermost layer of the BP-MSaaS architecture is the Web Layer, which implements the portal through which the user interacts with the platform. The portal is intended to be a one stop shop that provides an easy-to-use interface to the platform and its offered services. Underneath the Web Layer, the Middleware Layer provides the access point to the M&S services that are made available through the following components of the underlying cloud infrastructure: User Identification and Authorization: component providing authentication services.
Model Repository Manager: component providing services for storing/retrieving models to/from the cloud-based repository.
Modeling Service Manager: component providing modeling services for creating, editing and storing models. In order to make the architecture flexible enough to fit the needs of several application domains and modeling languages, modeling services can be automatically generated from a domain metamodel that defines a given modeling language, by use of transformation services.
Transformation Service Manager: component providing services for storing, retrieving and executing model transformations on the cloud infrastructure. Model transformations are used to generate models (e.g., simulation models from BP models), by applying model-to-model transformations, as well as to generate the code that implements executable services (e.g., modeling services or simulation services), by applying model-to-text transformations.
Simulation Service Manager: component in charge of managing the required tasks needed to first select a simulation service image available in the cloud repository and then deploy and execute it on the cloud infrastructure.
Presentation Service Manager: component providing services for the visual rendering of simulation results. Docker Infrastructure: component providing services for interacting with the underlying cloud infrastructure by use of the Docker containerization approach. transformation or edited by use of a modeling service, configurations data and service images. The executable code that implements a M&S service (i.e., a simulation, transformation or modeling service) is stored in the repository, along with the required library and runtime, as an executable image, which has to be deployed to a given container for execution.
Computation Service: to ease the execution of services on the computational nodes provided by the cloud infrastructure, the architecture exploits virtual containers, e.g., Docker [5], within which service images are deployed and executed.
Docker plays a relevant role in the proposed architecture as it provides the containerization feature enabling easy deployment of service implementations. The Docker architecture is founded on three main concepts: images, containers and registry. An image is a stand-alone and executable package that includes everything needed to run a piece of software (e.g., code, runtime, libraries, etc.). A container is a runtime instance of an image when actually executed. Finally, the registry, referred to as Docker Hub, is a repository of images. Docker also includes volumes, or directories stored outside of the containers filesystem, which are used to hold data that can persist even after container termination, thus providing data persistence to stateless Docker images.
Regarding the availabity of modeling, transformation and simulation services, it is worth noting that the following services have been already provided as standalone applications that exploit QVT [27], the standard provided by the OMG’s Model Driven Architecture (MDA) [28] for the specification of model-to-model transformations, and Acceleo [29], the modelto-text transformation language compliant to the MOFM2T MDA standard [30]:
Simulation Services: jEQN [31] and eBPMN [32] domain specific languages for enabling the simulation of EQN (Extended Queueing Networks) and BPMN models, respectively; Model-to-Text Transformation Services: Acceleobased transformations to generate jEQN code from EQN models [2] and eBPMN code from PyBPMN models, being PyBPMN the performance-oriented BPMN extension proposed in [33] and [34]. Moreover, an Ecore to Javascript and HTML Acceleo-based transformation to generate modeling services from domain metamodels specified in terms of Ecore models [35]; Model-to-Model Transformation Services: a QVTbased transformation to generate EQN models from UML activity diagrams [2] and a QVT-based transformation to generate PyBPMN models from BPMN models with text annotations specifying the related performance-oriented characterization [3].
The portal of the BP-MSaaS platform relies on the underlying cloud infrastructure layer, which provides the following services:
Storage Service: the cloud-based repository stores the several data and entities used by M&S services: models, which can be generated by a model Development activities are currently in progress to finalize the porting of such applications as microservices into the BPMSaaS platform. In this respect, the flexible and modular approach at the basis of the BP-MSaaS platform design and the availability of several modeling, transformation and simulation services make it possible to effectively use the BP-MSaaS Web Layer r e y a L e r a w e l d d i
M r e y a L e r u t c u tr s a fr n I d u o l C
WAN Infrastructure Portal e c i v r e S r iton age raom anM f s n a r
T
Docker y tr o iso re lepdeR aganM o M e c i platform in various application domains, by selecting the relevant set of services.
Finally, it is worth noting that the proposed architecture still provides the opportunity to make use of remote SOAP/REST Web services offered through a conventional WAN.
The next Section provides an operational view of the proposed approach by giving the behavioral specification of the BP-MSaaS platform.
Figure 2 shows the UML use case diagram that identifies relevant actors with related responsibilities. The specification of most important use cases is given later on in this Section.
Service Provider is the actor in charge of developing the several services available in the cloud, e.g., the above-mentioned eBPMN simulation service or the UMLto-EQN transformation service. The Service Provider also manages the service catalog by use of the Modeling Service Manager component provided by the BP-MSaaS platform, to deploy/delete services onto/from the cloud repository.
End User is the actor carrying out the simulationbased analysis of BPs. In this respect, it is in charge of configuring and executing a simulation experiment by selecting the simulation model and the required services from the catalog. The Simulation Expert supports the End User and is specifically involved in the definition of simulation models, which can be built from scratch or obtained by executing an automated model transformation (in such a case, models obtained by use of automated model transformations may need to be manually refined and configured).
Domain Expert is the actor that builds the domain models taken as input by model transformations used to generate other models (e.g., model-to-model transformations to generate simulation models) or the code that implements a service (e.g., a model-to-text transformation to generate a modeling service providing a Javascript-based visual editor). Finally, the System Administrator is in charge of managing the BP-MSaaS platform and the underlying cloud-based infrastructure.
The UML sequence diagram shown in Figure 3 provides the specification of the Build simulation experiment use case.
Service Provider End User Domain Expert System Administrator OPT
User
Login request Login confirmation Brose model catalog
REF
Model [Simulation model must be generated] Browse transformation catalog
REF Simulation model Configure(params) Execute(Simulation model)
REF Simulation results Build service Deploy service Delete service Browse models and services catalog Build simulation experimnent
«extends» Select and execute transformations Build simulation model «includes» Run simulation «includes» «includes» Select models Manage Cloud infrastructure «includes» Build domain model Store model Manage MSaaS portal <<UI>> Web Client <<Portal>> MSaas Portal <<Portal>> Docker Interface <<DockerRegistry>> Repository <<DockerEngine>> Execution Environment
Login() Login confirmation Select Models Select and Execute Transformations Run Simulation updateModel(params) updateModel(params)
Browse transformation
catalog [Simulation service available in the cloud] Transf images list Select transf image(s) <<UI>> Web Client Dock<<eProIrntatle>>rface getTransformations() Transf images list Transf images list getImage(imageList) getFromRepo(imageList)
Transf image(s) Build executable transformation set Start(tr_engine,model) Simulation model getFromRepo(tr_images) A set of transformation images is selected among the list of available transformations DockerSearch(tr_images) Transf images list DockerPul (imageList) Transformation image(s) DockerRun(tr_engine) Execute(tr_engine, model) The transformation engine (tr_engine) executes the set of selected transformations, in order to generate the simulation model from the input model Simulation model
< Portal> Docker Interface < DRoeckpeorRseitgoisrtyry>
< ServiceProvider> Simulation Service Simulation model Simulation model User Execute(SimulationModel) Execute(SimulationModel) OPT [Simulation service available in the cloud] getFromRepo(images) Engine image list Select engine Engine images list Engine images list getImage(imagelName) Simulation results [Simulation service available in WAN] Simulation results Simulation results Simulation results Fig.
Use case diagram of the
BP-MSaaS platfor
Fig. the
Build sim ulation experim ent use case. and includes
and
Run simulation use use case cases.
The Select model use case is first activated to browse the catalog and select a model, which may be at various levels of abstraction. In other words, if the selected model is a domain model or an abstract BP model, then the appropriate model transformations have to be selected and executed in order to generate an executable simulation model, as specified by the optional fragment in the diagram that refers to the Select and execute transformations sequence diagram. Differently, if an executable simulation model is the one selected by the user, then it is directly given as input to the appropriate simulation engine.
Figure 4 shows the sequence diagram specifying the Select and execute transformations use case, which is optionally executed when the selected model is not in executable format and thus one or more model transformations are to be carried Simulation Expert «extends» «extends»
Specification of the
Select and execute transformations use case.
Specification of the
Run sim ulation use case. out. In the considered scenario, the logged user browses the catalog of available model transformations and selects the set of required ones. It should be noted that the type and the number of the required transformation is strictly related to the type of the input model and the simulation engine which executes the simulation. As an example, an input UML activity diagram to be analyzed by use of an EQN model may require two different model transformations: a modelto-model transformation, to generate an abstract EQN model from the input UML activity diagram, and a model-to-text transformation to map the EQN model to an executable code ready to be taken as input by the EQN simulation engine. Differently, an input BPMN model may only require a single model-to-text transformation to directly generate the simulation implementation code for the available BP simulation engine. The required engine(s) are eventually deployed to the Docker container that executes the transformation(s) yielding the executable simulation model as output.
Finally, Figure 5 shows the sequence diagram specifying the Run simulation use case. The identified scenario illustrates how the BP-MSaaS platform allows users to execute simulation services deployed in the cloud or available as conventional SOAP/REST Web services. In the former case, the scenario overlaps with the Select and execute transformations use case, and thus the selected engine is deployed to a Docker container to enable the execution of the simulation model. In the latter case, the simulation model is used as input parameter of the invoked remote simulation service.
getFromRepo(image) Engine image Build executable simulation Start(engine,model) Simulation results Build service request startSimulaton(model)
DockerSearch(images) Engine images list DockerPul (image) Engine image DockerRun(engine) Execute(engine, model)
Simulation results
Simulation results Fig. 6. BPMN model of the Checkout BP.
IV.
APPLICATION SCENARIO
The proposed BP-MSaaS platform is used in this section to carry out the simulation-based analysis of a BP dealing with the check out of a purchase order submitted to an e-commerce website.
E-commerce application scenarios, being closely related to Business to Business Integration (B2Bi), can be effectively improved by use of the BP-MSaaS platform. B2Bi tasks are challenging because of the heterogeneous and distributed nature of systems in the enterprise network environment. ERP, CRM and logistic department are always triggered by ecommerce processes. The BP-MSaaS platform contributes to facilitate communication among the disparate systems and to recognize their dependencies with each other, which ultimately enables business process automation.
The model of the BP under study can be either retrieved from a catalog of available models or directly edited by use of a modeling service. Let us assume the BP is specified as a BPMN model built by use of a visual editor provided by the modeling service illustrated in [35]. Figure 3 shows the BPMN visual editor along with the BPMN model of the considered example.
The BPMN model has to be transformed into a simulation-oriented implementation and thus a model-to-text transformation is executed to take as input the BPMN model and yield as output the corresponding simulation code. The image of a transformation service is first searched and pulled from Docker Hub. Then, it is deployed and executed in a Docker container in order to get the desired simulation code. Such activities are transparently accomplished and supported by the BP-MSaaS platform.
Specifically, from the set of available services (as summarized in Section III-A) the BPMN-to-eBPMN Modelto-text Transformation Service is selected and executed in the example application scenario. The transformation takes as input the BPMN model and yields as output the simulation code specified in eBPMN [32], a BP simuation language built on top of the SimArch discrete-event simulation architecture [31].
Finally, the image of a simulation service is selected and executed, as specified by the sequence diagram in Figure 5. In this respect, an eBPMN-compliant simulation service is deployed and executed.
The Docker image which implements the eBPMN simulation engine is based on the official tomcat image availabe at https://hub.docker.com/ /tomcat/, a minimal runtime environment that provides Linux OS, Oracle JRE 8 and the Apache Tomcat 8.0 application service, which deploys the ebpmnService war file into the webapps folder.
Similarly to the generation of the eBPMN code, the BPMSaaS platform transparently provides what needed to build and execute the Docker container, as well as to package the war file that contains the executable simulation code.
Finally, an eBPMN-based presentation service is executed to retrieve the simulation results and provide them in visual form to the user.
As stated in Section I, the BP-MSaaS platform can be extended to cover additional application domains, which can be easily served by selecting a different set of M&S services. As an example, the performance analysis of a software system specified in UML can be enacted by selecting a UML-compliant modeling service, a different simulation service (e.g., the available jEQN simulation service) and an appropriate set of transformation services (e.g., the available
The paper has illustrated the proposed BP-MSaaS platform, which enables the effective adoption of M&S-based approaches throughout the BP lifecycle. Specifically, the BPMSaaS architecture exploits a cloud infrastructure and the Docker containerization architecture to build a flexible serviceoriented platform which provides the facilitated selection, deployment and execution of M&S services that support the business analyst in the specification and simulation-based analysis of BPs. The adoption of model-driven development approaches allows to ease the automated development of services and configuration files, in order to hide the complexity of simulation modeling and implementation activities, as well as platform management activities.
An application scenario dealing with an e-commerce BP has been illustrated to show a concrete example of how the BP-MSaaS platform can be effectively used to support and automate the selection, deployment and execution of modeling, transformation and simulation services.
The prototype Docker-based implementation of the BPMSaaS platform has been used to assess the feasibility and the effectiveness of the proposed architecture. Ongoing work includes the development of a full-featured implementation of the BP-MSaaS platform, by porting already existing applications as microservices into the platform, as well as by adding simulation engines that cover application domains not limited to the analysis of BPs.