Research On Self-adaptive Service Composition Optimization Approach Considering Alliance Relation

Thanks to the development of emerging computing modes such as cloud computing,more and more enterprises package their business functions as web services and open them to users on the cloud platform.Users can request services on the cloud platform according to their own needs,which is very convenient and simple.The phenomenon of bundling within service providers and cooperation among service providers is increasing day by day.It’s called Alliance Relation(AR),which will lead to the deviation of Quality of Service(Qo S).At the same time,the external environment of service is changing rapidly,which makes it difficult for the service composition system to obtain high credibility guarantee,which also directly affects the user experience and evaluation,attracting close attention from academia and industry.Therefore,it’s more and more important to fully consider the role of Alliance Relation and enable the service composition system to own the ability of self-adaptation.However,in the existing research on service composition,a large part does not consider the AR between services,but only considers the Qo S attributes of services themselves.In addition,the literature considering AR also faces the dilemma of insufficient consideration of AR types,the AR search space is too large,which leads to low solution efficiency.Moreover,there is a lack of relevant research on adaptive adjustment of service composition which take account of the dynamic impact of the external environment on Qo S.At the same time,inaccurate assessment of the long-term overall benefits of the system also leads to a lengthy adjustment process.To solve the above problems,firstly,we propose an alliance-aware service composition optimization approach with quotient space granulation.Based on granulation theory,according to the service granularity from fine to coarse,we construct the component granularity layer,function granularity layer,business granularity layer and quotient granularity layer,we divide the service granularity based on the quotient space theory.Then,the compatible AR are aggregated in the quotient space to form coarse-grained AR.On this basis,an AR index graph is created and a new graph search method is designed to query the optimal matching AR.According to the multi-objective function,NSGA-III algorithm is applied to optimize multiple Qo S dimensions at the same time,and finally a series of non-inferior service composition schemes can be obtained for users to choose.The approach can be widely used in all service AR types,and has a better optimizing each objective and higher computing efficiency.Then we continue to consider the continuous change of workload and gradually design the adaptive adjustment method of service composition to perceive the external environment.We first established an SLA double constraint framework for the dynamic Qo S attributes,delay and utilization,defining local constraints as hard constraints and global constraints as soft constraints.The local constraint is defined as hard constraint,while the global constraint is defined as soft constraint,in which the dynamic improvement of system benefits caused by AR is clarified.Once a service violates local constraints at a certain time point,TARCH time series model is used to predict the workload,which tracks the historical workload information to predict the future change trend.Then,a technical debt-aware model combined with the SLA framework is proposed to provide a reasonable basis for the adjustment of the service.Finally,we locate bad components based on the above processing,and quickly select the best composition service to replace them,which can maintain the stable operation of the business process as long as possible.This approach not only achieves better long-term global optimization effect for the business process system,but also makes the adaptive adjustment of service composition more sustainable.The main contributions of this paper are as follows.Firstly,aiming at the problem of incomplete coverage of AR in real scenes,three types of AR are considered,including adjacent AR,mixed AR and Superimposed AR.We expand the AR types of existing literature.Secondly,aiming at the problem of large composition search space,based on quotient space theory,we combine service granulation with alliance granulation,which improves the efficiency of service composition optimization considering AR.Thirdly,in view of the insufficient consideration of the long-term overall benefits of the service composition system in the external environment with dynamic workload,a self-adaptive method considering the dynamic impact of AR is proposed,which realizes better long-term global optimization effect and stronger stability.To sum up,this paper systematically considers the impact of AR on service quality,accelerates the query speed of optimal matching association,and improves the multi-objective optimization effect of service composition.Meanwhile,it accurately evaluates the long-term benefits of the service composition in a dynamic environment and maintain the stable operation of the business process as long as possible.