Deadlock prevention for service orchestration via controlled Petri nets

https://doi.org/10.1016/j.jpdc.2018.09.010Get rights and content

Highlights

  • •

    A controller-based abnormity prevention method for service collaboration is designed.

  • •

    A controlled reachability graph is proposed with structure and reachability analyses.

  • •

    A maximally permissive state feedback control policy to prevent abnormity is proposed.

  • •

    An optimal controller for deadlock prevention of service choreography is constructed.

Abstract

In multi-party Web service composition, the non-local choice mismatch is one of the most important problems. To solve it, the reachability graph-based method is widely adopted to check deadlock-freeness by generating adaptors. However, this method is inefficient due to the neglect of future deadlock states and requirements of many possible interactions. This work proposes an abnormity prevention strategy and an optimal controller for service collaboration. To overcome drawbacks of previous studies, this work describes service choreography by using service workflow nets. Moreover, this work formulates a novel reachability graph by combining structures and reachability analysis. To present abnormity, a maximally permissive state feedback control policy is then proposed. Furthermore, to avoid deadlocks in service orchestration, this work constructs an optimal controller for administrators of service composition. Finally, experiments demonstrate the advantage of the proposed method via a realistic example.

Introduction

Internet brings popular applications and computing paradigms, e.g., cloud computing, Web 3.0, mobile Internet, big data, Internet of Things and social networking [4]. The characteristics of Internet including openness, dynamics, and evolution, help us develop new technologies that evolve to effectively cope with fast changes in runtime contexts and user demands [9], [35]. The research about future Internet contributes to a distributed computing environment where a growing number of interactions are completed by using workflow service invocations. In Internet-scale and service-based workflow ecosystems, dynamic coordination and automatic composition for service workflow systems play an important role in realizing this concept [8], [43].

Existing service composition mechanisms typically realize the service orchestration according to Web Services Business Process Execution Language (WS-BPEL) [39], and they need a centralized approach to compose multiple services into a larger application [44]. It is shown that orchestration works well in static contexts with predefined services. However, these assumptions cannot well adapt to the Internet-based workflow ecosystem where different service workflow providers and consumers change their requirements, and therefore it is difficult to coordinate them with a centralized method.

On the contrary, service interaction specification languages, e.g., Web Services Choreography Description Language (WS-CDL) [48] and Web Service Choreography Interface (WSCI) [47], provide a decentralized method where a looser way is provided to realize service composition by determining participants and message protocols between them. Service choreographies present peer-to-peer message exchanges among participant services from a global perspective. Nevertheless, orchestration-based systems typically consider the following problems:

Firstly, a realizability check is essential to check if each implemented participant realizes orchestration provided that it conforms to the orchestration role specifying its expected behavior;

Secondly, a conformance check is essential to check if the global interaction of many services meets the orchestration [5]; and

Thirdly, web service orchestrations might fail to work well because inconsistent semantics can be buried in their interfaces and protocols among systems of different participants.

To solve these problems, many studies have proposed different methods [7], [10], [15]. However, most of these methods focus on direct composition between services. Partial compatibility is common in real-life web service interfaces and protocols. It means that functions of two or more web services are complementary, but their interfaces and protocols contain mismatches [17], [25]. Thus, they cannot be immediately composed. Besides, partial compatibility may cause problems including mutual waiting mismatch [38], unspecified reception mismatch [57] and non-local choice mismatch [33], [55]. This work aims to solve the non-local choice mismatch problem caused by partial compatibility. It can be formulated as: local choices of two services are different, and this causes deadlock [56], as shown in Fig. 1.

In Fig. 1, a service makes a local choice after it sends a message a (a!). For example, it sends a message b (b!) or c (c!). At the same time, its partner also needs to make a local choice, and waits for the message b (b?) or the message c (c?) after its partner receives message a (a?). If a service and its partner make the same choices, i.e., its partner is expecting to receive y while it is sending b, no problem arises. However, if its partner is expecting to receive c while it is sending b, a non-local choice mismatch occurs.

Enforcing partially compatible services to work with each other without any deadlock is critical and valuable to the reuse of services. In the case of non-local choice mismatches, it is assumed that the administrator of service composition establishes an optimal enforcement controller in the message channel for service orchestration that frees programmers from specifying subtle and hard-to-identify deadlocks. Therefore, programmers can focus on business process logic and depend on the optimal controller to provide a deadlock-free execution.

Main contributions of this work are summarized as follows:

  • (1)

    this work proposes a controlled Petri net-based method, i.e., controlled composition-of-service workflow net (C-CSWN) model, to avoid non-local choice mismatch in service orchestrations;

  • (2)

    this work establishes a controlled reduced reachability graph (CRRG) to efficiently determine deadlocks; and

  • (3)

    this work designs an optimal controller to realize an optimal control policy for incompatible service orchestration.

The rest of the work is organized as follows. A motivating scenario for compatibility analysis is presented in Section 2. A controlled composition of service workflow net is proposed to model service orchestration in Section 3. The compatibility analysis is implemented in Section 4 and a control policy is proposed to prevent the deadlock-prone orchestration. The method is proposed to build the controller in Section 5. The related work is presented in Section 6, and Section 7 concludes the work.

Section snippets

Motivating scenario

To illustrate the problem of non-local choice mismatch in service collaboration, the following example from supply chain domain is cited from [33], [55]. It is assumed that three web services expect to exchange messages and each has its own internal business process, i.e., Clients (C), Flight Booking (F) and Third-Party Checkout (T). They are developed and deployed in different Internet-based partners. Creating orchestration services is able to collaborate existing Internet-scale services. For

Interaction Petri nets

Similar to [3], [13], [14], [42], [50], [54], we model an Interaction Service Workflow Net (ISWN) for a business process as follows.

Definition 1 ISWN

An interaction service workflow net is an extended Petri net ISWN=P,T,F, where

  • (1)

    P is a finite set of state places. P=PI∪PM, where

    • (a)

      PI is a set of internal places;

    • (b)

      PM is a set of message places; and

    • (c)

      PI∩PM=∅;

  • (2)

    T is a finite set of transitions, P∩T=∅; and

  • (3)

    F⊆P×T∪T×P is a set of directed arcs connecting state places and transitions.

An ISWN is pure if it has no self-loops,

Compatibility analysis and control policy

This section shows the compatibility analysis of service orchestration. Besides, a solution is proposed based on an optimal control policy to prevent behavioral incompatibility in CSWN.

Optimal controller generation

Section 4.3 adopts CRRG to check existence of a controller among multiple ISWNs. This section further presents the approach to build it if a controller exists. Based on Theorem 2, CSWN is compatible, i.e., controller ℂ exists. To solve the above behavioral incompatibility problem, this work presents an algorithm by adding more controllers to CSWN as follows:

By adopting the proposed approach, this work concludes that CSWN, the interaction of ISWN1, ISWN2 and ISWN3, is compatible via controller ℂ

Related work

To analyze the compatibility of service orchestration, several formal semantics have been adopted, such as graphs [59], message sequence charts (MSC) [16], ontology [8], finite state machines (FSM) [6], process algebra (PA) [36], martin type theory (MTT) [58], and Petri nets (PN) [23], [32], [37], [52]. In particular, Petri net models have been widely adopted to describe service orchestration, from languages like BPEL to interaction Petri nets [13]. Therefore, it is appropriate for modeling and

Conclusions and future work

In principle, multiple web services provide complementary functions, but their interaction patterns do not fit each other exactly. To tackle this challenge, we have proposed a novel way to synthesize behavioral orchestration based on the optimal controller. Firstly, we model a service orchestration as a controlled Petri net. Base on this model, this work analyzes the incompatibility of multiple services with a reduced reachability analysis technique. Then, this work constructs a controlled

Acknowledgments

We thank all the reviewers for their insightful comments. This work was supported in part by the Grants from the National Natural Science Foundation of China (No. 61703011 and 61802015).

Jing Bi received the Ph.D. degree from Northeastern University, Shenyang, China, in 2011. She is currently an Associate Professor with the School of Software Engineering, Beijing University of Technology in the Faculty of Information Technology, Beijing, China. She is currently also a visiting research scholar with the Department of Electrical and Computer Engineering, New Jersey Institute of Technology (NJIT), Newark, NJ. From 2013 to 2015 she was a Postdoc researcher at Department of

References (59)

  • BertolinoA. et al.

    Software engineering for internet computing: internetware and beyond

    IEEE Softw.

    (2015)
  • BlakeM.B. et al.

    Composition as a service

    IEEE Internet Comput.

    (2010)
  • CarboneM. et al.

    Structured communication-centred programming for web services

    Proc. 16th European Symposium on Programming (ESOP) as part of the European Joint Conferences on Software Theory and Practice (ETAPS), Braga, Portugal

    (2007)
  • CoffmanE.G. et al.

    System deadlocks

    Comput. Surveys

    (1971)
  • DeckerG. et al.

    Local enforceability in interaction petri nets

  • FantiM.P. et al.

    Deadlock control methods in automated manufacturing systems

    IEEE Trans. Syst. Man Cybern. A

    (2004)
  • FosterH. et al.

    Tool support for model-based engineering of web service compositions

  • GarlanD. et al.

    Architectural mismatch: why reuse is still so hard

    IEEE Softw.

    (2009)
  • GhaffariA. et al.

    Design of a live and maximally permissive Petri net controller using the theory of regions

    IEEE Trans. Robot. Autom.

    (2003)
  • GodefroidP. et al.

    Using partial orders for the efficient verification of deadlock freedom and safety properties

    Form. Methods Syst. Des.

    (1993)
  • HanL. et al.

    Efficient optimal deadlock control of flexible manufacturing systems

    IET Control Theory Appl.

    (2016)
  • HollowayL.E. et al.

    Synthesis of feedback control logic for a class of controlled Petri nets

    IEEE Trans. Automat. Control

    (1990)
  • HollowayL.E. et al.

    A survey of Petri net methods for controlled discrete event systems

    Discr. Event Dyn. Syst.

    (1997)
  • HuH. et al.

    Maximally permissive distributed control of large scale automated manufacturing systems modeled with Petri nets

    IEEE Trans. Control Syst. Technol.

    (2015)
  • IchikawaA. et al.

    Analysis and control of discrete event systems represented by Petri nets

  • KongdenfhaW. et al.

    Mismatch patterns and adaptation aspects: a foundation for rapid development of web service adapters

    IEEE Trans. Serv. Comput.

    (2009)
  • KroghB.H.

    Controlled petri nets and maximally permissive feedback logic

  • LiY. et al.

    Control of vector discrete-event systesm I—the base model

    IEEE Trans. Automat. Control

    (1993)
  • LiY. et al.

    Control of vector discrete-event systesm II—controller synthesis

    IEEE Trans. Automat. Control

    (1994)

    • Cited by (5)

    Jing Bi received the Ph.D. degree from Northeastern University, Shenyang, China, in 2011. She is currently an Associate Professor with the School of Software Engineering, Beijing University of Technology in the Faculty of Information Technology, Beijing, China. She is currently also a visiting research scholar with the Department of Electrical and Computer Engineering, New Jersey Institute of Technology (NJIT), Newark, NJ. From 2013 to 2015 she was a Postdoc researcher at Department of Automation, Tsinghua University, Beijing, China. From 2009 to 2010, she participated in research on cloud computing at the IBM China Research Laboratory. Her research interests include distributed computing, cloud computing, large-scale data analytics, machine learning and performance optimization. Dr. Bi was the recipient of the 2009 IBM Ph.D. Fellowship Award.

    Haitao Yuan received the B.S. and M.S. degrees in Software Engineering from Northeastern University, Shenyang, China, in 2010 and 2012, respectively, and the Ph.D. degree in Control Science and Engineering from Beihang University, Beijing, China in 2016. He is currently an Assistant Professor with the School of Software Engineering, Beijing Jiaotong University, Beijing, China. He is currently also a visiting research scholar with the Department of Electrical and Computer Engineering, New Jersey Institute of Technology (NJIT), Newark, NJ under the financial support from the China Scholarship Council (CSC). He was also a visiting doctoral student with NJIT in 2015. His research interests include cloud computing, data center, software-defined networking, optimization algorithms, and big data. Dr. Yuan was the recipient of the 2011 Google Excellence Scholarship.

    Wei Tan received the Ph.D. degree from the Department of Automation, Tsinghua University, Beijing, China. He is currently a Research Staff Member at the IBM T. J. Watson Research Center, Yorktown Heights, NY, USA. From 2008 to 2010, he was a Researcher at the Computation Institute, University of Chicago, and Argonne National Laboratory, USA. At that time, he was the technical lead of the caBIG workflow system. His research interests include GPU acceleration, distributed computing, business and scientific workflows, and Petri nets. He has published over 60 journal and conference papers, and a monograph “Business and Scientific Workflows: A Web Service-Oriented Approach” (272 pages, Wiley–IEEE Press). His research results have been incorporated into IBM big data products and patent portfolios. Dr. Tan is a member of the Association for Computing Machinery (ACM). He received the Best Paper Award from ACM/IEEE ccGrid 2015, the IBM Outstanding Technical Achievement Award (2014 ), the Best Student Paper Award from IEEE ICWS 2014, the Best Paper Award from IEEE SCC 2011, the Pacesetter Award from Argonne National Laboratory (2010), and the caBIG Teamwork Award from the National Institute of Health (2008). He is an Associate Editor of the IEEE TRANSACTIONS ON AUTOMATION SCIENCE AND ENGINEERING.

    View full text
    © 2018 Elsevier Inc. All rights reserved.