Research On Temporal Correctness Of Business Workflows In The Cloud

E-business and E-government applications have been constantly developing in the last two decades.To satisfy the requirements of business process automation and execution efficiency,business workflow has been widely used in many business domains,and many business workflow systems have sprung up.A notable characteristic of business workflow application is that there are usually a large number of workflow instances running in a parallel and distributed fashion,where each instance represents a unique user or system request.Given the requirement in the processing of concurrent business requests,cloud environment that can provide cost-effective scalable resources is regarded as an ideal host environment for running large number of workflow instances.Therefore,more and more organizations are looking for ways to move their business process management environment to the Cloud.Typical business cloud workflow system includes Amazon Simple Workflow and IBM BPM on Cloud in industry and SwinFlow-Cloud and CloudBus in academia.For the applications in business scenario,the most important and primary attribute is logical correctness.Business requests must be handled in a correct logic so that correct results will be delivered to the system or users.While as a kind of typical real-time system,the computation correctness of business workflow applications not only depends upon the logical correctness of the computation,but also upon the temporal correctness,namely whether the response time of business request is consistent with time constraints or temporal specifications in Service Level Agreement(SLA).Failing to deliver requested results in time may lead to not only the deterioration of user satisfaction,but also the expiration of results,which could result in significant financial loss.Therefore,the rate that the concurrent workflow instances are completed within time constraints(on-time completion rate for short)is one of the most critical Quality of Service(QoS)dimensions for business workflow system in the cloud.However,because of the dynamic nature and uncertainties that exist during the running of workflows in the cloud,workflow instances are prone to temporal violations,which have an impact on the on-time completion of workflow instances.Here "temporal violation" means an intermediate violation of time constrains that can be locally to ensure overall timely completion.Furthermore,resource sharing and temporal dependencies among workflow activities accelerate the propagation of temporal violations in the workflow system,which will seriously jeopardize the timely completion of massive concurrent business workflow instances.In general,ensuring the temporal correctness is a big challenge for business cloud workflows mainly for the following three reasons:(1)the massive concurrency of business workflow instances makes their temporal states difficult to be monitored at runtime;(2)due to the dynamic nature of cloud infrastructure and propagation effect of response delays during the running of parallel workflow instances,temporal behaviors of workflow instances are difficult to be measured and verified;(3)business workflow instances are normally executed with a short duration,to make sure the business requests can be responded within constrained time,the detected temporal violations need to be handled timely and effectively.Although,various approaches for workflow verification have been proposed,they are not applicable to the temporal correctness verification in business scenario.To tackle the challenge above,this paper carries out a series research work on the temporal correctness of business workflow in the cloud(business cloud workflow for short),the aim is to ensure that real on-time completion rate of business cloud workflow instances achieves the target.It should be noted that "temporal" in this paper means time-related information that are reset by designer at workflow build-time stage,such as average execution time of workflow activity and overall response time of workflow instance."Temporal consistency" means the satisfaction of time constraints during the runtime of workflow instances,namely the consistency between workflow runtime temporal state and the temporal QoS(on-time completion rate in this paper)that is set at workflow build time.Specifically,the following three majority studies are conducted in this paper:(1)Static prediction of workflow temporal violationsLifecycle of a workflow consists of workflow build time stage and runtime stage.This paper presents an epidemic model based temporal violation prediction strategy.The proposed strategy is conducted at workflow build time state,which can estimate the number of workflow activities that may violate local temporal constraints at workflow runtime and determine the number of temporal violations that must be handled to achieve the target on-time completion rate.Prediction results can serve as essential information for the optimization of workflow configuration(such as time constraints and execution container setting),temporal violation prevention and handling.To the best of our knowledge,this is the first attempt to predict cloud workflow temporal violations at the workflow build-time stage.(2)Propagation-aware temporal consistency verification for business cloudworkflowsIn order to monitor and verify runtime temporal consistency state of business workflow instances running in the cloud,this paper presents a propagation-aware temporal consistency verification strategy.Compared with the existing verification approaches,the proposed strategy performs better in both verification efficiency and effectiveness.Specifically,to reduce verification overhead,instead of using response time of workflow activities to monitor every single workflow instance,workflow throughput is employed as the performance measurement for describing temporal behavior of parallel workflow instances.To achieve a higher verification accuracy,a new throughput consistency model is proposed for the temporal consistency monitoring and verification purpose.This model considers the propagation effect of time delays in the cloud workflow environment and can produce a much more accurate verdict of temporal consistency for a large number of parallel cloud workflow instances.Based on the model,a novel consistency verification model is presented to verify workflow runtime temporal consistency and timely detect temporal violations.(3)Adaptive workflow temporal violation handlingTo achieve on-time completion of time-constrained business cloud workflows,the detected temporal violations need to be timely handled.For this purpose,this paper presents an adaptive workflow temporal violation handling strategy,which adjusts workflow temporal behavior by recruiting extra resources to accelerate the execution progress of workflow instances.The proposed temporal violation handling strategy are designed from the perspective of resources.First,this strategy locates the exact cloud services where temporal violations occur.Then the number of extra resources that should be recruited for temporal violation is analyzed.Finally,an adaptive method is designed to dynamically determine the lifecycle of recruited resources.In general,to ensure target on-time completion of business workflow instances running in the cloud,this paper investigate three key research problem on workflow temporal correctness,and the corresponding solutions are also presented.The first study is to predict temporal violations at workflow build time,and the last two studies is implemented at workflow runtime to monitor temporal behaviors,verify temporal consistency and handling temporal violations.To demonstrate the effectiveness and efficiency,the proposed strategies in this paper are implemented in a prototype cloud business workflow system SwinFlow-Cloud.