A Hypermedia Middleware for Process Enactment and Adaptation

Roy Oberhauser
2016 International Journal of Software Engineering and Its Applications  
Dynamic business process management incorporates the capabilities needed to deal with the fast-paced change inherent in today's business processes. While process-aware information systems (PAIS) face an increasing challenge to dynamically adapt running processes to context changes, support for such dynamic adaptation during process enactment remains limited. Furthermore, while the use of web service APIs for enterprise application integration (EAI) and cloud accessibility has garnered broad
more » ... unity support, current PAIS lack a standardized API and often require use of their proprietary clients or APIs. In this paper, we explore the use of hypermedia as the engine of application state (HATEOAS) for process enactment and process adaptation. Our approach Adapting Processes via Processes using hypermedia (AProProh) situates a HATEOAS-based middleware between RESTful clients and heterogeneous PAIS that can dynamically guide a PAIS-agnostic process client in the navigation, enactment, and adaptation of process instances. Dynamically generated hypermedia enables clients to dynamically apply valid adaptations and adjust their process navigation dynamically to a changed process model while supporting long-running operational requests. A case study based on a prototype realization shows the viability of the approach for supporting dynamic process navigation and process adaptation and characterizes its performance. 54 Copyright ⓒ 2016 SERSC lack of standardization and accessibility. Although apparent standards such as BPMN [4] for process modeling notation and BPEL [5] as a process execution language seem common, nevertheless interchange issues exist [6] . Actual internal process model formats and execution languages, which often preceded the standards, are often PAIS-specific and thus exhibit some differences and further possibilities. Moreover, a standardized API to simplify the integration of PAIS from various vendors is not evident or widely supported. As an alternative for such integration, hypermedia as the engine of application state (HATEOAS), a constraint of the Representational State Transfer (REST) application architecture [7] , supports dynamic navigation of REST APIs by a REST client based on hypermedia knowledge. REST can be seen in some ways as a type of runtime workflow, yet the application of HATEOAS for EAI in currently available adaptive PAIS for process navigation and adaptation has not been adequately explored. Since the business environment has become much more dynamic, there is increasing pressure for the capability to dynamically and to some extent automatically adapt processes for changing contextual conditions, also known as dynamic business process management (dBPM). However, currently available PAIS rarely support correctness and soundness guarantees for runtime process adaptation [8] , and when then the focus is typically on supporting manual process changes by a human actor [9] . Any recurring process modifications are known as workflow control-flow patterns, change patterns, or adaptation patterns [9] . dBPM will necessitate automated process adaptation, and these will need a viable process schema and a process enactment adaptation capability as well as a way to model the adaptations. To address both the dBPM and the integration challenge, this paper, a revised and expanded version of [10], elaborates an approach and hypermedia middleware for dynamic process enactment and adaptation called Adapting Processes via Processes using hypermedia (AProProh). It is a hypermedia extension of the original AProPro (Adapting Processes via Processes) [11] , which models and enacts adaptations imperatively, practically expressing and maintaining adaptations in an intuitive and sustainable manner for the dBPM lifecycle. Towards a standardized web API mechanism for PAIS integration, this HATEOAS-PAIS middleware includes support for long-running activities, processes, or process adaptations. The evaluation is based on a case study that shows its viability with a prototype and assesses the middleware's performance characteristics with different REST implementations. The paper is organized as follows: Section 2 discusses related work, which is followed by a description of the solution approach. In Section 4, details of the prototype realization are provided. An evaluation based on a case study is given in Section 5, followed by the conclusion. Related Work As to process adaptation, various approaches support the manual or automated adaptation of processes. Case handling approaches [12] utilize a case metaphor, deemphasize activities, and are data-driven [9] . Declarative approaches, such as DECLARE [13] support the constraint-based composition, execution, and adaptation of workflows. Agent-based approaches include Agentwork [14] , which applies predefined change operations using rules, and [15], where a belief-desire-intention agent utilizes a goal-oriented BPMN modeling language extension. Aspect-oriented approaches include AO4BPMN [16], which requires a language extension. Variant approaches include: Provop [17], which supports schema variants with pre-configured adaptations to a base process schema; and vBPMN [18] , which extends BPMN with fragment-based adaptations via the R2ML rule language. rBPMN [19] also interweaves BPMN and R2ML. Automated planning and exception-driven adaptation approaches include SmartPM [20] , which utilizes artificial intelligence, procedural, and declarative elements. 55 In contrast, the AProPro adaptation approach supports runtime adaptation and is imperative without utilizing a case metaphor. It does not require language extensions or paradigms such as rules, declarative elements, or intelligent agents. Work related to hypermedia in combination with processes includes HATEOAS in combination with BPMS or PAIS. [21] involves combining RESTful with BPM, but does not mention of HATEOAS constraints or hypermedia, nor are changes to process models undertaken or performance impacts discussed. RESTful and BPEL work includes [22] , which focused primarily on the composition and integration of services. Similarly, BPMashup [23] focuses on process-centric compositions of RESTful web services without addressing hypermedia. RESTfulBP [24] and CPEE [25] do not mention HATEOAS or hypermedia. RESTful with BPMN [26] extends BPMN notation with graphical syntax and semantics for REST support, but without directly addressing hypermedia or HATEOAS. Various commercial BPMS vendors support REST interfaces, yet to our knowledge HATEOAS is not explicitly mentioned.
doi:10.14257/ijseia.2016.10.6.05 fatcat:36idji6zcnbczideh3emfntpoa