BPMN 1.2 tutorial

Draft

This article or section is currently under construction

In principle, someone is working on it and there should be a better version in a not so distant future.
If you want to modify this page, please discuss it with the person working on it (see the "history")

Introduction

This article provides a short overview of BPMN 1.2. BPMN is a so-called flowcharting notation language for defining a business process in terms of connected activities, events and gateways. BPMN is a Business Process Diagram (BPD) language that includes functionalities found in workflow modeling languages used in computer science, e.g. UML activity diagrams and UML use case diagram. However, BPMN is more complex and seems to be better suited to model workflows that involve humans.

We suggest reading a tutorial by example first, e.g. try BPMN by example from Bizagi. A list of other tutorials is also available in the BPMN overview article.

The "Business" word in standards like BPMN, BPEL, etc. reflects the fact that this language was specifically developed by and for the business community. BPMN can be used to model any "bureaucratic" process that involves both humans and computers (including web services). BPMN (an particularly the upcoming 2.0 version) may have interesting potentiel for education, in particular for modeling and implementing activity-based learning designs such as CSCL scripts.

See also: Business process modeling (for background information), BPMN (overview and links including software) and BPMN 2 tutorial

Overview of BPMN 1.1/1.2 (2008/9)

According to the BPMN 1.2 specification, “The primary goal of BPMN is to provide a notation that is readily understandable by all business users, from the business analysts that create the initial drafts of the processes, to the technical developers responsible for implementing the technology that will perform those processes, and finally, to the business people who will manage and monitor those processes. Thus, BPMN creates a standardized bridge for the gap between the business process design and process implementation. Another goal, but no less important, is to ensure that XML languages designed for the execution of business processes, such as BPEL4WS (Business Process Execution Language for Web Services), can be visualized with a business-oriented notation.”

BPMN 1.2 is a so-called notation language, i.e. it allows to "draw" the design of a business process, i.e. the coordination of human and computer workflows in an organization. In computer science, its closest kin is the UML activity diagram. Like some UML languages, BPMN is a kind of flow-charting language. However, the specification includes an informal annex that defines possible mapping to BPEL (more precisely the BPEL4WS version), a process execution language. In addition, vendors also could implement translators to any sort of execution environment. BPMN 2 (to be finalized by the end of 2010) defines its own serialization format.

BPMN was created with business execution languages and web services in mind and it is based on a mathematical foundation, the Pi-Calculus. However, in practice. BPMN 1.2 only has "execution" and "web services" in mind, i.e. they are not formally "included" in the specification. In practice this means that you either can use BPMN as descriptive modeling language only or face vendor lock-in if you are looking for a full modeling-implementation-execution pipeline.

BPMN core elements

BPMN is a fairly complex and rich language. According to the 1.2 specification, BPMN has 11 core elements for defining a business process and that can be grouped in four categories:

(a) Flow Objects

define the behavior of a business process and are the most important constituent elements of process diagrams.

Types: (1) Event (start, intermediate, end), (2) Activity (task, sub-process and process), (3) Gateway (various)

BPMN 1.2 - Flow objects

(b) Artifacts (Artefacts)

are used to provide extra information about the process

Types: (4) Data Object, (5) Group, (6) Annotation

BPMN 1.2 - Artifacts

(c) Swimlanes

group elements into processes performed by entities

Types: (7) Pool, (8) Lane

BPMN 1.2 - Swimlanes

(d) Connecting Objects

connect flow objects and artifacts (above) to each other

Types: (9) Sequence Flow, (10) Message Flow, (11) Association.

A simple example

With the help of these connecting objects, process diagrams can be structured in three ways:

1. Sequence flows define an order of activities performed in a process. Each flow has a single source (a start event, activity or gateway) and a single target (a end event, activity of gateway). Elements of a sequence flow are connected with solid arrows.
2. Message flows either connect two pools or flow objects within these pools. As opposed to sequence flows, they cannot connect objects within the same pool. Message flows are represented with dashed arrows.
3. Associations associate flow objects with information and artifacts and are drawn with dotted lines.

The following figure implements a simple workflow with two processes that are coordinated by message flows, i.e. dotted arrows. Each process must have a start (the green circle) and an end (the red circle). It then should include at least an activity, in our case 1 or two simple tasks. Flow objects within a same process (e.g. start event, task and end event) are connected by solid arrows representing sequence flows.

BPMN 1.2 - Connecting objects

Message and sequence flows are usually drawn orthogonally, i.e. sequence goes in one direction (either top->down or left->right) and messages in the other as the picture above shows.

The BPMN 1.2 extended set

Most of the eleven core elements can by typed, e.g. an event can be either of type, start, intermediate or end event. These sub-types then can be further typed, e.g. events may have an associated trigger. As an example, the start event may be a message start event meaning that the process is triggered (started) after receiving a message from an other participant.

Below is a simple example of a pedagogical scenario. Tutor hands out a reading assignment, learner reads the book and then creates a summary with a online tool. The teacher will receive a message and then can provide some feedback. The student then has a revise it and the teacher will grade it after some deadline.

BPMN 1.2 - Simple reading/summarizing assignment example. (Made with ARIS Express 2.2)

Disclaimer: This is my first BPMN drawing, not sure that is correct - DKS/July 5/2010

As you can see in the figure and as we shall explain later, tasks also can be typed. In this simple scenario, we use "manual tasks" (user reading a book), computer-supported "user tasks" (using the hand symbol) and a message task.

Notice: I am not sure how data objects should be used (in the model the book and summary are linked to more than one task, this may not work in some vendor frameworks as we found out. In addition, data objects are not a necessity, since data also can (invisibly/implicitly) attached to tasks, flows and events. Anyhow, it seems to us that dataflow is somewhat underspecified in BPMN or in other words, BPMN focusses on the organization of task and dataflow is something that has to be dealt with "behind" the notation language. Daniel K. Schneider 16:35, 20 July 2010 (UTC)

We will introduce the various components of the extended BPMN set in the following sections. But before, we shortly look some syntactic and excution issues.

BPMN process syntax and semantics

It is important to understand that there are many restrictions on how BPMN elements may be combined. The specification is (mostly easy enough to understand, but it is very difficult to get global use case pictures, i.e. what combinations of elements to use for a given kind of process. In addition, since BPMN is a pure notation language, a modeler has the liberty to design a process a various levels of abstraction/precision. However, once he uses a commercial framework that integrates BPMN with an data modeling and process execution framework this doesn't hold true anymore. Finally, there often several alternatives for modeling a same pattern which is a bit confusing.

Overally speaking, the rules for connecting elements of a business process are fairly simple. Look at the previous figure for an examples that show simple processes that includes two participants (i.e. two sequence flows) and a message flow between these participants.

Sequence flow rules

• In principle, any flow object of start event, intermediate event, task, sub-process, gateway can connect to intermediate event, task, sub-process, gateway and end event. Or the other way round: no object can connect to a start event, and end events cannot connect other events.
• Objects within subprocesses cannot connect to objects outside this process.
• Sequence flows cannot cross a pool boundary, use messages or signals for this purpose.

Put simply: As long as flow objects sit inside a pool you can try to connect these in almost any way. However, there are many exceptions to this general rule and that are detailed in the specification. In principle, a BPMN editor ought to refuse illegal connections...

Message flow rules

• pools, subprocesses, tasks, intermediate message events and end message events can connect to pools, subprocesses, tasks, startmessage events and intermediate message events. Within this list, only start and event events behave differently: Start message events can only receive a message, but not send a message. End message events can only send messages but not receive messages. Let's recall that start events start a process and end events end a process.

Drawing rules for connecting lines and arrows

There are no limitations with respect to drawing, exit and entry points of connecting lines can be anywhere on the border of an object. However, it is obvious that a flow diagram should be as simple as possible (mostly straight lines and little overlaps).

Data modeling and execution

BPMN 1.x does not specify any language that might be used as input for an workflow execution engine, it just specifies and informal way to map BPMN to BPEL4WS. BPMN doesn't even specify a representation (serialization) language for the drawings, meaning that you can't export/import a drawing from one editor to another.

Put simply, PBMN drawings must be neither portable nor executable, although several vendor-specific solutions do exist.

Some vendors provide tools for exporting/importing their BPMN 1.x diagrams to their own workflow/web services framework. Some can import/export to XPDL (a serialization language for process diagrams) and BPEL (a BPM execution language), however many authors claim that BPEL is not the holy grail for BPM.

Another difficulty relates to the sometimes not very clear connecting objects semantics of BPMN 1.2 or maybe implementations we played with were incomplete. It seems that the question whether BPMN Semantics is "Vague" or Just Hard to Code? remains open. Finally, it seems that a given problem can be modeled in many different ways and we wonder how this could affect potential execution models.

Below, as an example, we sketch out a simple BPM modeling/implementation process as it is implemented in Bizagi Studio and its built-in development runtime engine.

(1) We first did another take of the simple reading/summarizing assignment example shown above. We used a single pool, i.e. considered that this scenario was to be modeled as a single integrated workflow, i.e. on process. Since Bizagi Process modeler did not accept more than a input/output association flows couple for a single data object, we then had to replicate the same data object (representing the summary a learner had to write and which we meant to be a wiki page) into three separate objects representing three versions of the same data object.

Simple reading/summarizing assignment example (Made with Bizagi Process Modeler 1.5.1)

This valid model according to Bizagi process modeler did not validate with the Bizagi Studio BPM Suite, since the latter seem to require that data objects must be associated with a single task. I.e. we got 4 error messages complaining that "the data object must be associated with an activity". Actually they were associations, but two instead of one. When will CS people learn about error messaging ? .... Anyhow, we modified the model like this:

Simple reading/summarizing assignment example (Bizagi Studio)

(2) In a next step (once a model is imported into a BPM(N) engine, one must model the data. The whole process is represented as one or more entities with attributes. This is IMHO the most difficult step (harder than modeling the process itself .... more later).

(3) After data is modeled, one then can define user input forms, i.e. the user interface that will be used by the execution engine.

(4) Finally the model then can be loaded and executed as "case" by an execution engine.

Activities

An activity stands for the work that an organization (e.g. a company) performs. An activity can be atomic or non-atomic (compound). The types of activities that are a part of a Process Model are:

• Tasks, represented by rounded rectangles plus (optional) additional markers;
• Subprocesses, represented by rounded rectangles. They can be collapsed (annotated with a + sign) or not;
• Processes, represented with swimlanes, i.e. either pools with lanes or simple pools. Both can be collapsed.

Simple task markers

Since BPMN is a notation language, the language used is often oriented towards drawing. Therefore "markers" are used to visually distinguish sub-classes of task objects.

BPMN task markers

An "atomic" (simple) task is a simple activity performed within an organization. A task correspond to an activity step in the process flow that cannot/should not be broken further down. Generally speaking, a task is either performed by a human, a computer application or both (i.e. a user using an application)

BPMN defines three basic kinds of tasks that are visually defined with markers: loops, multiple instance loops and compensations. Markers are graphically standardized and must be placed at the bottom of the rectangle. The compensation marker can be used together with loop or multiple instance. Let's now shortly describe these "markers":

(1) Loop

Loops are repetitive tasks and include a test that defines if a task is to be repeated after each cycle of the loop (p. 54/p.65). May be used together with the compensation and multiple instance markers.

Standard loop attributes include:

(a) LoopCondition - Expression

Expressions are defined with respect to a scripting language, i.e. the modeler must both provide a string defining a scripting language and a valid expression.

(b) LoopCounter - Integer

Counts the cycles and may be used in the condition

(c) LoopMaximum - Integer

Defines maximum loops

(d) TestTime - "Before"/"After"

Defines either a while or until type of loop, i.e. the loop condition is tested before or after each cycle

(2) Multiple instance loops

Are a variant of loops and define for instance whether loop instances will perform in parallel or sequentially. Multiple instances also allow to define gateways after parallel path, i.e. where the "token" will go.

(3) Compensation

A compensation activity is special sicne it does not follow the normal sequence flow rules (p. 130). A compensation is one of the three types of "undo" activities: overwrite any changes (revert), do nothing, or invoke an activity that undoes the effects. The latter is called compensation. E.g. a student may turn in a late assignment and will get 10% off. If later, the student will show a medical certificate, then he will get the 10% reimbursed. However, and this is why it's not a simple undo, his assignment is still "done".

Simple task types

BPMN 1.2 task types

In addition to these main task markers representing a kind of task subclass, BPMN defines so-called task types. There is no standard graphical representation for these. Implementors can choose their own and also are allowed to add additional types (something that we certainly would need for education).

Roughly speaking, the most important task types specify which ones are automatic, IT supported or carried out manually. BPMN 1.2 defines the following seven task types:

(1) Service task

A service task provides some sort of service like a web service or another automated application. In other words, these tasks are entirely automatic, i.e. executed by a computer.

(2) Receive task

A Receive Task is designed to wait for a message to arrive from an external participant (relative to the Business Process). The receive task:

• can't have an incoming sequence flow
• the incoming sequence flow has a Start Event
• it is completed once the message is received

(3) Send task

Send Task is designed to send a message to an external participant (in another pool, relative to the Business Process)

• The task is completed once the message is sent.

(4) User task

The user task is a typical “workflow” task where a human performer performs the Task with the assistance of a software application.

• A user task is scheduled through a task list manager of some sort

(5) Script task

A Script Task is executed by a business process engine. The modeler or implementer defines a script in a language that the engine can interpret.

• When the Task is ready to start, the engine will execute the script and the task is completed at the completion of the script.

(6) Manual task

A Manual Task is expected to be performed without the aid of any software (like a business process execution engine, a web service or any other kind).

(7) Reference task

When modeler may want to reference another task that has been defined (exactly the same behavior attributes)

As we already explained above, there exist some rules for both sequence flow connections and message flow connections. In addition:

• A task may have multiple sequence incoming flows although this is not recommended, i.e. one should use a gateway before the task to control flows in different paths
• ... (to do)

Sub-processes

A sub-process is a compound activity within a flow of other activities (p. 56). It has several purposes and comes two variants and serveral sub-variants.

Example of BPMN 1.2 subprocesses

An expanded sub-process can be used to (a) associate this compound activity with exception handling, (b) with compensation handling, and finally (c) implement parallel activites in a less cluttered way than using gateways.

A collapsed sub-process may be used to hide detail within a complex process. Collabsed (only) sub-processes then can then include a combination of up to three Loop Marker, Parallel Marker, Compensation Marker, and Ad Hoc Marker. Loop cannot be used with parallel (i.e. multiple instance).

Both collabsed and expanded sub-processes can be transactions as for example defined in WS-transaction, a OASIS (formerly BEA/MS/IBM) web services specification.

Swimlanes

A sequence flow always happens in a so-called pool or in a lane that is part of a pool. Both simple pools or pools with lanes are called swimlanes. Each swimlane usually represent a different independant participant that intervenes in the process.

In the business world, a pool represents an organization and a lane a unit within an organization. In that sense, lanes are used like the ones in UML activity diagrams, i.e. they represent "who does what". But a modeler is free to use pools and lanes for any other sort of partitioning, e.g. a lance could represent an organizational or technical function or a system.

In order to model educational learning designs (pedagogical scenarios), we might use a single pool to represent a strongly integrated workflow such as the ones that could be implemented with a workflow systems like LAMS and separated pools if the workflow is loosely organized such as in our C3MS project-based learning model using portalware or decoupled web 2.0 services.

Gateways

Types of BPMN 1.2 gateways

“Gateways are modeling elements that are used to control how Sequence Flow interact as they converge and diverge within a Process. If the flow does not need to be controlled, then a Gateway is not needed. The term “Gateway” implies that there is a gating mechanism that either allows or disallows passage through the Gateway--that is, as Tokens arrive at a Gateway, they can be merged together on input and/or split apart on output as the Gateway mechanisms are invoked.” (BPMN 1.2 specification, p 70).

Gateways are graphically represented as diamonds, like in many other flow chart languages. However, BPMN Gateways are used to represent many kinds of sequencing behavior: Decisions/branching (exclusive, inclusive, and complex), merging, forking, and joining as we shall see. Incoming and outgoing sequence flows can connect to any boundary position.

(more todo ...)

BPMN 1.2 event elements

An event is something that "happens" during a process and it will affect the flow of the process. It usually has either a cause or an impact or both. Events typically do not depend on the user, but on an external factor.

In more operational terms, one also could say that events will affect the sequence of the timing of process activities.

Start, intermediate and end event

There are three main types of events: start events, intermediate events and end events.

• Events are represented as small circles and fills must be either white (no color) or use a clear color.
• Start events have a fine stroke, intermediate events a double fine stroke, and end events a fat stroke.
• Inside the circle there can be shape that helps identifying a trigger. As you can see below, the various event types can be associated with various triggers, but not all triggers are available for each of the three start, intermediate and end event.

BPMN 1.x symbols, Credits: BPMN 1.1 stencils by F.Puhlmann/A. Grosskopf

Start Events

Start events indicate the initiation of a process and are usually associated with a trigger. The five types of triggers defined by v. 1.1 are represented with graphical markers.

Plain (generic) start events

do not specify any trigger

Message start events

start when a message is received from an other participant

Timer start event

indicate a given time/date for starting

Conditional start events

a process is triggered when certain business conditions become true.

Signal start events

A signal arrives that has been broadcasted from another process

Multiple start events

Defines multiple ways of triggering a process.

Intermediate events

There are 10 types of intermediate events in BPMN 1.2. All of these can respond to (catch) a trigger and some can set off (throw) a trigger. Catching events (only) also can be directly attached to the boundary of an activity.

Events are drawn as double circle. Catching event markers are drawn with strokes and throwing events are drawn filled.

Simple ("none") intermediate event

Events in the main flow of the process that are not further specified.

Message (catching and throwing) intermediate events

When catching, a message arrives from a participant and triggers the Event. If the process was waiting for the message it then can continue. Messages also can be used for exception handling, i.e. the process will continue differently.

When throwing in normal flow, a message is sent to another participant.

Summarized, messages are used to receive/send between 2 participants A and B.

Timer (catching) intermediate event

A set time/date will trigger the event, i.e. it inserts a delay. For example, it could be used to model a deadline for turning in an assignment.

Error (catching) intermediate event

Only can be used in the boundary of an object. Can either catch a named error or all errors if none is specified

Cancel (catching) intermediate event

This cancel event must be used in the boundary of a transaction subprocess. It can be triggered either by a message or by a cancel end event in the subprocess.

Compensation (catching and throwing) intermediate events

When catching, a compensation activity attached to this event will be triggered.

In normal flow, when throwing an event is defined, this event will be triggered, else "compensation" is broadcasted to all activities in the process.

Conditional (catching) intermediate event

Event is triggered when a condition becomes true

Link (catching and throwing) intermediate event

Links connect two sections of a process (at the same level), e.g. to implement looping, goto or simply move to a "next page". There can be multiple source, but only one target link.

Signal (catching and throwing) intermediate event

Signals can be sent and received across process levels, pools and even diagrams. As opposed to messages, targets cannot be identified, i.e. there is always a specific source (throwing) but specific target. In other words, a signal is like a flare. Catching events must be attached to the boundary of an activity.

Multiple (catching and throwing) intermediate event

In normal flow, a multiple event can either catch or throw various events. When attached to an activity it can only catch.

End events

Simple ("none") end event

showing that and where a process ends

Message end event

indicates that a message is sent to a participant when the end of the process is reached

Error end event

an error code should be generated. This code then can be caught by an error catching intermediate event in the parent. If there is none, then the system also can handle this.

Cancel end event

is only used in transaction sub-processes and will signal that this transaction should be cancelled.

Compensation end event

can either compensate an activity that is identified. Otherwise all activities in the process can be subject to compensation if they have a compensation intermediate event attached.

Signal end event

a signal is generated. It will be broadcasted to any process (i.e. signal catching events) that can receive it. A signal is different from a message. The latter always has a source and a target.

Terminate end event

triggers an immediate end to the process and all its activities. The process is ended without compensation of event handling.

Multiple end event

multiple consequences of ending the Process.

Artifacts

Data objects

Data objects do not directly affect sequence or message flows. They provide information about what the process does, i.e. how data, documents and and sort of other objects are used and updated during the process.

PBMN doesn't model artifacts themselves, this task is left to vendors according to the specification (p 93.). We found the specification to very vague with respect to these objects and indeed various modeling tools we tried (Aris Express, Bizagi Process Modeler, Bizagi Studio) won't allow to use data objects in the same way - Daniel K. Schneider 17:58, 21 July 2010 (UTC).

Data objects are generally associated with flow objects (typically a task). In that case they can represent input or output with directional arrows.

They also can be attached to sequence and message flows in order to show what data are sent from one activity to another. In the case of a message flow, the data object represent the "content" (payload) of the message itself.

The data object includes a name, a state and various modeler-definable properties. The latter then can be mapped to a data model in an execution engine.

Some design patterns for connecting objects

(to do)

Examples

Model of a discussion

BPMN discussion model. Source: Erik Wilde, Business Process Execution Language (BPEL), Slides, UC Berkeley iSchool

Email voting process

:BPMN-E-Mail Voting Process. Source: Wikipedia/unknown

Model of surgery programming

The process model presented by Roja et al. (2008) shows at a high level abstraction the programmed surgical patient (PSP) process and the activities that are carried out when a patient is admitted into the hospital for the accomplishment of the surgical treatment of a disease that has been clinically diagnosed previously. This model is a result of a joint work between hospital and BPMN expert and it has been obtained through a reiterative and incremental method.

Programmed surgical patient: Source: Roja (2008) et al. 10.1186/1746-1596-3-S1-S22.

“The modelling of the programmed surgical patient process and its subprocesses has allowed to us to prepare an understandable model for the involved health professionals and make the communication of processes easier. Additionally, modelling allows early detection and correction of errors. This work is an essential previous step for further analysis and improvements in healthcare processes, including the adoption of information technology standards.” (Roja et al. (2008))

A BPMN 1 meta model

A more difficult to understand meta-model (i.e. an inofficial UML class diagram of BPMN was published by WSPER ("whisper").

BPMN 1.0 Meta model. Source: WSPER

Links and bibliography

See BPMN for more links, books, software, etc. Below just a few links.

• Dive into BPMN
• Modeling Workflow Patterns through a Control-flow perspective using BPMN and the BPM Modeler BizAgi by Marcus Goetz, Institute of Applied Informatics and Formal Description Methods University Karlsruhe (TH).
• BPMN_Tutorial.pdf by Martin Owen and Jog Raj, Popkin Software
• BPMN by example and BPMN Tutorial/Overview (Bizagi)