Elsevier

Future Generation Computer Systems

Volume 86, September 2018, Pages 464-479
Future Generation Computer Systems

A sufficient and necessary temporal violation handling point selection strategy in cloud workflow

https://doi.org/10.1016/j.future.2018.03.056Get rights and content

Highlights

  • Development of a temporal consistency model for workflow activities in the queuing systems.

  • Time redundancy is considered into the execution process of cloud business workflows.

  • Proof of sufficiency and necessity for the temporal violation handling point selection strategy.

Abstract

To deliver high QoS (quality of service) for business process participants, workflow temporal verification is conducted to provide satisfactory on-time completion rate of business process in the cloud. Temporal violation handling is the last task in a typical workflow temporal verification framework to deal with detected time delays. However, there are very few existing studies regarding temporal violation handling for cloud business workflows. In this paper, queuing theory is first employed to simulate time features of parallel workflow instances. Then, propagation effect based temporal consistency model for business workflows and temporal consistency model for workflow activities in the same queuing system are presented respectively. Finally, a promising temporal violation handling point selection strategy for cloud business workflows is proposed and proved to satisfy the property of sufficiency and necessity. Compared with other representative strategies, experimental results show that our novel handling point selection strategy can reduce the monitoring and handling cost while maintaining the target on-time completion rate agreed between users and service providers.

Introduction

Business processes are the series of interactions between businesses and their customers, vendors and other related partners. A business workflow is an instance of well-defined business process that are often repeated as part of a standard enterprise operations [1]. In general, these business workflows can be managed by a workflow system and executed in a fully or semi-automatic fashion. With the fast trend of globalization, increasing popularity of e-commerce and e-government systems requires effective and efficient processing of massive service requests within a deadline, which has become ever more important [2].

In many business domains, tens of thousands or even millions of business workflows happen frequently and the total number of workflow instances is extremely large on a daily basis. A business process can be very simple in some cases, for example, buying a newspaper or paying two dollars for a cup of coffee only needs a few steps to complete the whole process [1]. Some processes are more complex, for example, purchasing an item with credit card involves a series of tasks before the purchase can be completed, and there could be million or even more workflow instances in a securities exchange corporation which need to be handled during the peak time of the stock market [3]. No matter how simple or complex the processes are, a common characteristic is that the amount of these workflow instances is enormous. Some of these business processes are often time constrained [4] that should be completed before fixed time points. For instance, securities exchange in the stock market is a typical complex instance-intensive business process which involves a large number of instances between different organizations [5]. For example, the clearing process and money transfers in the stock market may need to be finished before 3:00 am each workday. This can be referred as the time to completion (TTC) issue. It is defined by the span from the start to the completion without iteration [6]. As the time span is strictly constrained and thus any time delays may cause lower quality of service (QoS). Failures of on-time completion for money transfer in stock market may cause huge financial losses as it could result in the failure of making deals, which is definitely a disastrous situation in the stock market. So on-time completion (OTC) of time-constrained large number of workflow instances is critical for delivering satisfactory QoS in today’s e-business and e-government process systems.

Different from traditional scientific workflows, instance-intensive business workflows usually involve a large number of workflow instances [6]. Each instance normally has a few steps [7] and some steps of the workflow instance can be executed concurrently. Business processes are typical instance-intensive workflows which are characterized by a large number of concurrent and often relatively simple workflow instances [8]. Business processes normally have a specific deadline and need to achieve the target on-time completion rate for certain business objectives [9]. As shown in Fig. 1, temporal constraints are serving as a type of service level agreement (SLA) as part of QoS requirement [10]. There are two types of temporal constraints that we need to consider in business workflows. The first type is global constraint which applies to the entire workflow process. The second type is local constraint which applies to each individual workflow activity. For the reason that large number of business processes are dynamically changing all the time, it is very necessary to employ cloud computing resources to deal with the functional requirement of business workflows. Many existing studies [[11], [12], [13], [14]] have demonstrated that cloud computing is a good facility for offering virtualized, on-demand self-service, broad network access, resource pooling, rapid elasticity and measured service [15]. So cloud computing is an ideal hosting environment for running a batch of parallel business workflow instances [[4], [13]] such as securities exchange. It can supply scalable and sufficient resources to complete business workflows within the deadline of each workday.

As shown in Fig. 1, the four basic components for temporal verification of business workflows are temporal constraints setting, temporal checkpoint selection, temporal consistency verification and temporal violation handling in a workflow life cycle. To facilitate the use of cloud resources, three big challenges should be resolved at the moment to improve on-time completion rate for business workflows. First, where do we set checkpoints so as to monitor the execution process? Too much monitoring may cause huge waste while too less may cause many temporal violations undetected. Hence checkpoints need to be selected at appropriate time points. Second, how do we verify the execution consistency state of large number of business workflow instances? Temporal violations often inevitably occur in workflow applications due to the uncertainty and dynamic nature of cloud environment. Hence the temporal consistency verification is a key issue when temporal violations are detected at checkpoints. Third, when do we handle workflow temporal violation? Time redundancy exists in workflow instances when there is no time deficit. It is normally unnecessary to trigger temporal violation handling for a minor time deficit as there is a high probability that it will be automatically compensated for by the time redundancy of the subsequent activities. Hence, whether temporal violation handling (TVH) is necessary or not depends on different situations.

Based on the above challenges, we need a trade-off solution to guarantee both the overall throughput (OT) of business workflow instances and each individual response time (IRT) of workflow activities. First, queuing model is employed so that large number of business processes can be executed in parallel. Then a propagation-aware temporal consistency model is applied to account for the effect of time delay propagation in queuing system. Afterward, an overall temporal verification framework is presented for parallel business workflows, which verifies the system temporal consistency state at selected checkpoints by the temporal consistency model. Finally, a new minimum time redundancy-based handling point selection strategy is proposed to deal with temporal violations in business processes with minimum number of handling points.

The overall workflow temporal verification framework based on queuing theory can guarantee on-time completion for a batch of business workflows in a constrained time. More importantly, the novel temporal violation handling point selection strategy has been proved to satisfy the property of sufficiency and necessity by two theorems, which can reduce the cost for monitoring the execution process remarkably.

To sum up, our significant contributions in this paper are as follows.

(1) For the first time, we develop a novel temporal consistency model for workflow activities in the same queuing system, which facilitate to verify the consistency state of workflow activities in the same queuing system.

(2) For the first time, time redundancy of completed workflow activities is considered in the execution process of a large number of parallel workflow instances.

(3) For the first time, we propose novel temporal violation handling point selection strategy for cloud business workflows and prove the property of sufficiency and necessity.

The remainder of the paper is organized as follows. Section 2 introduces some related work. Section 3 illustrates a motivating example of business process with constrained time. Section 4 proposes temporal consistency models based on queuing theory. Section 5 presents a novel temporal verification framework for cloud business workflows. Section 6 elaborates the detail experimental results. Section 7 concludes the paper and puts forward some future work.

Section snippets

Related work

In recent years, there are many works on temporal verification for a large number of cloud business workflows. Some traditional cloud workflow researches mainly focus on single process of managing and executing large-scale scientific computing issues in many complex e-science applications [[16], [17]]. Scientific workflow is often composed by a series of computation and data intensive tasks which are normally time-constrained such as the climate and astrophysics applications [[18], [19]].

Motivating business process in constrained time

Securities exchange is a typical instance-intensive business workflow which involves a large number of instances and each of them is a relatively simple workflow instance with only a few steps [7]. The steps of the workflow instance are always executed in certain order. There are two kinds of processes, including stock trading process and stock clearing process as shown in Fig. 2, Fig. 3.

Fig. 2 is the trading process of securities exchange, which is an online-mode process in stock market. There

Temporal consistency model based on queuing system

In this section, we mainly focus on temporal consistency model based on queuing theory. Firstly, queuing model is applied to simulate the execution process for a great deal of parallel workflow instances in Section 4.1. Secondly, some definitions associated with workflow throughput are given for designing the temporal consistency model in Section 4.2. Thirdly, propagation effect of time delays in workflow queuing system is briefly introduced in Section 4.3. Finally, temporal consistency model

A novel temporal verification framework for business workflows

Workflow temporal verification framework for business workflows is composed of four components, which are (1) temporal constraint setting, (2) temporal checkpoint selection, (3) temporal consistency verification and (4) temporal violation handling point selection respectively. In this section, since there has been many research on temporal constraint setting [[19], [20], [22]], we will elaborate the latter three in detail as they are directly related to this paper.

Evaluation

In this section, to evaluate the effectiveness and applicability of our proposed strategy for business workflows compared with some representative methods, comprehensive simulation experiments are implemented for a large amount of parallel workflow instances.

Conclusion and future work

Affected by the dynamics and uncertainty of cloud computing environment, a large number of parallel business processes are difficult to avoid temporal violations during the whole execution period. Therefore, workflow temporal verification is implemented to delivery desirable on-time completion rate for business workflows. In this paper, first, queuing theory is applied to simulate the queuing process regarding a large batch of parallel workflow instances. Then, propagation effect based temporal

Acknowledgments

The research is supported by the National Natural Science Foundation of China (Grant No. 61602005, 71601001), MOE Youth Project of Humanities and Social Sciences, China (No. 14YJCZH169), Natural Science Foundation of Anhui Province, China (1608085MF130, 1808085MF199), Natural Sciences Research Project in Universities of Anhui Province (KJ2018A0016), China, Anhui University Doctor Startup Fund, China .

Rongbin Xu received the Master’s degree from Anhui University, China, in 2007, and the Ph.D. degree from Anhui University, China, in 2015. He is currently a full lecturer in the School of Computer Science and Technology, Anhui University, China. He is a member of CCF and his current research interests include cloud computing, workflow systems, and social network.

References (30)

  • XuR. et al.

    Near-optimal dynamic priority scheduling strategy for instance-intensive business workflows in cloud computing

    Concurr. Comput.: Pract. Exper.

    (2017)
  • da SilvaR.F. et al.

    A characterization of workflow management systems for extreme-scale applications

    Future Gener. Comput. Syst.

    (2017)
  • WuZ. et al.

    A market-oriented hierarchical scheduling strategy in cloud workflow systems

    J. Supercomput.

    (2013)
  • JinT. et al.

    Refactor business process models with maximized parallelism

    IEEE Trans. Serv. Comput.

    (2016)
  • RoyS. et al.

    An empirical study of error patterns in industrial business process models

    IEEE Trans. Serv. Comput.

    (2014)
  • Cited by (22)

    • A collaborative scheduling method for cloud computing heterogeneous workflows based on deep reinforcement learning

      2023, Future Generation Computer Systems
      Citation Excerpt :

      Meeting the diversity of computing business demands under an elastic cloud architecture is an urgent issue that needs to be considered [1]. As an ordered collection of massive computing tasks, workflows usually appear in the form of directed acyclic graphs [2] and are widely executed with the growth of complex computing requirements in various application fields. The workflow usually contains a large number of task nodes, and the concurrent computing demand fluctuates greatly, especially in heterogeneous workflow combinations.

    • Improved Long Short-Term Memory based anomaly detection with concept drift adaptive method for supporting IoT services

      2020, Future Generation Computer Systems
      Citation Excerpt :

      Therefore, how to ensure high privacy of normal communication between different devices is a very important issue for smart home [7]. Because it serves as the security guarantee that users can enjoy the convenience brought by different IoT services, it can improve the quality of service (QoS) significantly [8]. In the IoT environment of smart city, the real status of communication is recorded and collected in a chronological order which indicates that there is a temporal relationship in each record.

    • Smart fog based workflow for traffic control networks

      2019, Future Generation Computer Systems
      Citation Excerpt :

      For example, it can be tailored to your Needs; all data is stored in one Location; reduced dependency on IT and so on. Yuan et al. and Xu et al. have presented algorithms and strategies to make cloud workflows more efficient [16] [17]. AI Technologies, especially in deep learning, which consists of various derivatives of artificial neural network like convolutional neural network (CNN), recurrent neural network (RNN) and so on, have been successfully applied in the fields such as speech recognition, image classification, machine translation [18].

    • A novel directional and non-local-convergent particle swarm optimization based workflow scheduling in cloud–edge environment

      2019, Future Generation Computer Systems
      Citation Excerpt :

      Based on the above motivating example, workflow application scheduling in cloud–edge environment will be defined as the formulated problem of assigning computing resources with different processing abilities to the tasks of workflow application, which can minimize the total completion time and cost of workflow application scheduling. A business workflow is an instance of well-defined business process that are often repeated as part of a standard enterprise operations [30,31], which can be considered as a view or representation of the real world from one step to another. In general, a workflow consists of a series repeatable patterns of business activities that can transform materials or provide services for business applications.

    • A comprehensive multi-objective approach of service selection for service processes with twofold restrictions

      2019, Future Generation Computer Systems
      Citation Excerpt :

      2) We will improve the algorithm by reference to Cooperative Coevolutionary and MapReduce framework [10,48], so as to increase the searching effectiveness and efficiency for big service selection. ( 3) Our approach does not concern the practical situations on the uncertainty of services, inveracious services and malicious evaluation [24,49]. Therefore, we plan to address some more practical approaches as further research items, such as investigating creditability evaluation of services and considering the heterogeneous QoS values with uncertainty, so as to improve the realistically and scalability of our approach.

    • Auto-Scaling Containerized Applications in Geo-Distributed Clouds

      2023, IEEE Transactions on Services Computing
    View all citing articles on Scopus

    Rongbin Xu received the Master’s degree from Anhui University, China, in 2007, and the Ph.D. degree from Anhui University, China, in 2015. He is currently a full lecturer in the School of Computer Science and Technology, Anhui University, China. He is a member of CCF and his current research interests include cloud computing, workflow systems, and social network.

    Yeguo Wang is currently a postgraduate in School of Computer Science and Technology, Anhui University, China. His research interests include workflow management systems, workflow scheduling and cloud computing.

    Haoyu Luo is currently a Ph.D. student in State Key Lab. of Software Engineering in Wuhan University, China. His research interests include workflow systems and cloud computing.

    Futian Wang received the Bachelor and Master degrees from Anhui University, Hefei, China in 2005 and 2009, respectively. He is currently a lecturer toward the Ph.D. degree in Anhui University. His research interests include software engineering, cloud computing and workflow systems.

    Ying Xie received the Master’s degree from Anhui University, China, in 2008, and the Ph.D. degree from Anhui University, China, in 2014. She is currently a full associate professor in the School of Computer Science and Technology, Anhui University, China. She is a member of CCF and her current research interests include cloud computing, machine learning, and social network.

    Xiao Liu received his Ph.D. degree in Computer Science and Software Engineering from the Faculty of Information and Communication Technologies at Swinburne University of Technology in 2011. He is currently a Senior Lecturer at School of Information Technology, Deakin University, Melbourne, Victoria, Australia. His research interests include workflow systems, cloud computing and social network.

    Yun Yang received his Ph.D. degree in Computer Science from the University of Queensland, Australia, in 1992. He is currently a Full Professor in Faculty of Science, Engineering and Technology at Swinburne University of Technology, Melbourne, Victoria, Australia. His current research interests include software technologies, cloud computing, workflow systems and service oriented computing.

    View full text