Modeling and Executing Batch Activities in Business Processes (Extended Abstract)

Keywords

• Luise Pufahl

Motivation

Organizations strive to provide useful products and services to their customers in an e cient manner. For reaching this, the underlying business processes need to be designed in an e ective and e cient manner. Business process management (BPM) provides methods and techniques for the documentation, digitalization, enactment and evaluation of business processes [ 19, 2 ]. The main artifact being produced thereby is a process model serving as a blueprint for several process instances; a process instance represents the execution of one business case [ 19 ].

A common assumption in BPM is that the process instances have an independent existence and are executed autonomously [ 16 ]. However, the consolidated execution of cases is a common phenomenon in operational business processes. For instance, in healthcare, it is more time-e cient to rstly collect a set of blood samples taken from patients to deliver them to the laboratory instead of bringing each one of them separately. Or, when calling a software service in a business process, bundling several service requests can help to reduce the number of service calls. In these use cases, batch processing [ 5 ] is applied. It allows business processes acting on a single item to bundle the execution of a group of process instances for particular activities to improve their performance.

Batch processing of products [ 7, 13 ] or services [ 8, 10 ] is deeply investigated in operations management, but an integration of its techniques in process models was not considered so far. In existing business process modeling languages, such as BPMN (Business Process Model and Notation) [ 11 ], and BPM systems, the explicit design and execution of batch activities is not yet supported. Also, the often referenced control ow patterns [ 1 ] do not discuss batch processing. Multiinstance activities [ 1 ] are presented where multiple instances are created and synchronized at the end of their execution, but they are still independently executed. Thus, batch processing is often organized manually by practitioners where the rules of the batch work are not documented and cannot be enforced.

Regarding the given problem statements, the doctoral thesis in [ 14 ] provides a comprehensive concept for integrating batch activities into business process models. The thesis followed the design science process [ 12 ] starting with a requirements analysis, developing and evaluating a designed artifact, and improving that in several cycles. In the remainder of this extended abstract, the contributions of the referenced thesis, its implications, and future research will be described. 2

Contributions

In the discussion of the related work, it was observed in the thesis that batch processing is considered in other domains, such as computer science and operations management. Whereas batch processing in computer science is used to e ciently process a large amount of data with no user interaction, discussing the design and implementation of such systems, in operations management batch processing is used to process similar products or groups of customers for being more e cient, studying the balance between reduced costs and increase in cycle time. Only a small set of research works in the BPM domain exists on the integration of batch processing in business process models. The presented solutions in those works focus on speci c scenarios and lack of a complete understanding of requirements. Therefore, the thesis started with a requirements analysis based on which the design objectives were set. Then, the main contribution was presented, the concept on how to explicitly represent batch activities in business process models and to execute them automatically. Aspects of exibility during the batch execution by di erent means (i.e. exible design of the batch activation rule, user involvement strategies, and exible batch con guration based on events) were presented and discussed. Additionally, the basic concept was extended to batch processing across multiple di erent process models. In the following, the main four contributions of this thesis are presented in detail: I. Requirements Framework In the referenced doctoral thesis, a requirements framework for integrating batch processing in business processes was presented. It was developed based on related work and complemented by requirements from collected industry examples, taken from di erent domains. The requirements framework gives insight into the aspects which need to be considered for developing a batch processing concept for business process models. Additionally, it fosters also the comparison of existing solutions. In this thesis, the requirements framework was used to structurally compare the requirements of the collected real-world scenarios whereby two preliminary types of batch activities { automated and user-involved batch activities { were identi ed. Further, it was used to set the design objectives, and to compare the developed batch activity concept to other related work.

II. Batch Activity Concept The main result of the thesis is the batch activity concept. The concept describes the syntax of a batch activity with its batch con guration parameters which need to be speci ed by process designers. Specifically, we present in details the groupedBy -parameter for grouping instances in speci c batches based on their data context and the batch activation rule being responsible for balancing the cost reductions with additional waiting time. Di erent types of activation rules were formalized, such as the threshold rule identi ed in operations management. Based on it, extensions were developed, such as the MinMaxRule to consider future instances. Further, it gives an operational semantics of the batch activity. It describes the life cycle of batch clusters (i.e. the actual representations of batch executions) and the interaction of batch clusters with process instances and the activity resources (i.e. task performers or services). The feasibility of the concept was shown by a prototypical implementation in Camunda { an existing, open-source BPM system. The application of the batch activity to di erent use cases shows that process cost can be reduced with acceptable or even positive in uence on the cycle time if a suitable batch activation rule is selected. For supporting automated as well as user-involved batch activities, di erent levels of user involvement and the way they might be realized were presented in this thesis.

III. Integration of Flexibility in Batch Activities The thesis showed that exibility aspects of batch activities are not discussed by existing related work in the BPM domain. In the work of this thesis, exibility for batch activities was provided by di erent means: First of all, batch activation rules, which are using Event-Condition-Action rules [ 4 ], can be exibly designed, such that also an immediate execution of special cases is allowed. The thesis presented here the so-called FastTrackRule. Further, the presented user involvement strategies allow task performers to react dynamically on changes and exceptions in the process environment (e.g. by starting batch clusters when needed, re-assigning instances to another batch cluster, or waiting for speci c future instances). Finally, the thesis presents a concept based on event processing techniques to allow a exible batch con guration. In this, batch adjustment rules are de ned by process designers which specify for which event type which type of batch clusters need to be adapted and how. The application of the batch adjustment rules to a healthcare scenario in a simulated environment showed that they help to compensate the losses caused by the exceptional behavior in this use case.

IV. Multi-process Batch Processing Concept This thesis could show that batch processing across several di erent business processes is useful and was not yet discussed by existing related work. The proposed multi-process batch processing concept allows a centrally de ned batch speci cation in an object lifecycle. Object lifecycles complement process models and describe allowed actions of business processes on data artifacts across the process-model boundaries. The basic concept of batch transitions was additionally extended in this thesis to multiply connected batch transitions (to allow also batching in process fragments) and to multiple similar batch transitions (to allow batching of activities having di erent data inputs but producing the same output). The requirement of optional batch processing in a multi-process setting was enabled by activating the batch processing only when similar activity instances are detected. Further, the concept includes a user approval where the task performer has to accept identi ed batches by the system. The feasibility of the concept was shown by a prototypical implementation in Camunda, an existing open-source BPM system.

Implications and Future Research

The results of the referenced thesis have relevant implications for practice and the BPM research. With regards to practice, it showed that batch work cannot be speci ed and automated with the current process modeling languages and technology. Therefore, a new process modeling element was provided, the batch activity that allows a speci cation of batch work with the help of several con guration items. Additional to that, this thesis presented a BPMN process simulator Scylla [ 15 ] extended with batch activities which can be used to validate a dened batch activity and its con guration to identify the most useful one. The extension of Camunda by the batch activity provides for BPM system providers an example how this new modeling element can be integrated into their system. Also, the given operational semantics in the thesis can support such e orts.

With regards to BPM research, the referenced doctoral thesis extended and advanced the rst attempts of batch activities (e.g. in [ 5, 17 ]) with a comprehensive requirement analysis and a concept for batch activities building on this. The proposed batch activity concept stimulated or in uenced further research, such as batch process mining (e.g. in [ 6 ]), new types of batch activities (e.g. in [ 9 ]), and that batching of instances is a relevant instance-spanning constraints [ 3 ]. The exibility concept for batch activities shows how runtime changes can be handled in a process with the help of event processing; thereby it contributes to the area of exible process management. The extension of batch activities to batching over several business processes showed that object life cycles are useful to de ne central aspects being relevant for several business processes.

Future work with regards to batch activities can be focused on the following directions: Batch activities are currently identi ed and con gured based on expert knowledge by the process designers. The developed BPMN process simulator Scylla being able to simulate business process with batch activities could be extended to give recommendations for an optimal batch activity con guration. Another possibility, if execution logs exist, is to use process mining techniques for an automated discovery of batch activities from those event logs as proposed by [ 6, 18 ]. However, those techniques are not capable to give recommendations for batch activities and to mine a complete batch con guration, such that more research work in this direction is needed. Further, an application of the batch processing concepts presented in this thesis (e.g., in a user study or in a technical action research [ 20 ]) will give more insights into the usability and usefulness of the presented concepts. These might also lead to new requirements which extend or detail the introduced requirements framework. With regards to exibility of batch activities, this thesis presented di erent means to support the batch activity adaptation, but the ability to allow variability of batch activities was not discussed yet. In future, a concept to allow variability of the batch activity congurations can be developed which might be supported by the batch processing discovery techniques. Going a step further, the current batch activity concept which is a design and implementation concept could be used as a basis for a dynamically batch processing concept where batches can be created at any activity if certain conditions are ful lled.

References

1. van der Aalst , W.M., ter Hofstede , A.H. , Kiepuszewski , B. , Barros , A.P. : Work ow patterns . Distributed and Parallel Databases 14 ( 1 ), 5 { 51 ( 2003 )

2. Dumas , M. , La Rosa , M. , Mendling , J. , Reijers , H.A. : Fundamentals of business process management , vol. 1 . Springer ( 2013 )

3. Fdhila , W. , Gall , M. , Rinderle-Ma , S. , Mangler , J. , Indiono , C. : Classi cation and formalization of instance-spanning constraints in process-driven applications . In: International Conference on Business Process Management (BPM) . pp. 348 { 364 . Springer ( 2016 )

4. Laliwala , Z. , Khosla , R. , Majumdar , P. , Chaudhary , S. : Semantic and rules based event-driven dynamic web services composition for automation of business processes . In: Services Computing Workshops (SCW) . pp. 175 { 182 . IEEE ( 2006 )

5. Liu , J. , Hu , J. : Dynamic batch processing in work ows: Model and implementation . Future Generation Computer Systems 23 ( 3 ), 338 { 347 ( 2007 )

6. Martin , N. , Swennen , M. , Depaire , B. , Jans , M. , Caris , A. , Vanhoof , K. : Retrieving batch organisation of work insights from event logs. Decision Support Systems ( 2017 )

7. Mathirajan , M. , Sivakumar , A.I.: A literature review, classi cation and simple meta-analysis on scheduling of batch processors in semiconductor . The International Journal of Advanced Manufacturing Technology 29 ( 9-10 ), 990 { 1001 ( 2006 )

8. Medhi , J.: Stochastic models in queueing theory . Academic Press ( 2002 )

9. Natschlager, C. , Bogl, A. , Geist , V. , Biro , M. : Optimizing resource utilization by combining activities across process instances . In: European Conference on Software Process Improvement . pp. 155 { 167 . Springer ( 2015 )

10. Neuts , M.: A general class of bulk queues with poisson input . The Annals of Mathematical Statistics 38 ( 3 ), 759 { 770 ( 1967 )

11. OMG: Business Process Model and Notation (BPMN ), V. 2 . 0 ( 2011 )

12. Pe ers, K. , Tuunanen , T. , Rothenberger , M.A. , Chatterjee , S.: A design science research methodology for information systems research . Journal of Management Information Systems 24 ( 3 ), 45 { 77 ( 2007 )

13. Potts , C.N. , Kovalyov , M.Y.: Scheduling with batching: A review . European journal of operational research 120(2) , 228 { 249 ( 2000 )

14. Pufahl , L. : Modeling and enacting batch activities in business processes . Ph.D. thesis , University of Potsdam ( 2018 )

15. Pufahl , L. , Wong , T.Y., Weske , M. : Design of an Extensible BPMN Process Simulator . In: Business Process Management Workshops ( 2017 )

16. Russell , N., van der Aalst , W.M. , Ter Hofstede , A.H. , Edmond , D. : Work ow resource patterns: Identi cation, representation and tool support . In: International Conference on Advanced Information Systems Engineering (CAiSE) . pp. 216 { 232 . Springer ( 2005 )

17. Sadiq , S. , Orlowska , M. , Sadiq , W. , Schulz , K. : When work ows will not deliver: The case of contradicting work practice . International Conference on Business Information Systems (BIS) 5 , 69 { 84 ( 2005 )

18. Wen , Y. , Chen , Z. , Liu , J. , Chen , J.: Mining batch processing work ow models from event logs . Concurrency and Computation: Practice and Experience 25 ( 13 ), 1928 { 1942 ( 2013 )

19. Weske , M. : Business Process Management: Concepts, Languages, Architectures . Second Edition . Springer ( 2012 )

20. Wieringa , R.J.: Design science methodology for information systems and software engineering. Springer ( 2014 )