Trusted Service Composition And Its Key Technologies In Cloud Environment

As a new and promising computing paradigm, cloud computing can provide users with scalable services (such as hardware, software, platform and virtualized applications) on demand by leveraging networking and virtualization techniques. In the past few years, with the rapid development of Internet and cloud computing technologies, the services in the cloud environment present some new features such as large scale, high complexity, dynamics and diverse types of failure. Meanwhile, the application requirements of users are becoming more complex and personalized. In such an environment, how to provide trusted service recommendation and composition efficiently with consideration of users’ personalized requirements, and how to ensure the reliable execution of users’ application, are the key issues of the research on trusted service application in cloud computing environment.In traditional services computing paradigm, there have been some researches on trusted services. However, in cloud computing paradigm, the scale of service resources becomes larger, the applications of users are more complex, and the network environment is more dynamic and unstable. All these new features of services bring challenges in researches on trusted service recommendation and trusted service composition:1) With the development of cloud computing technology, the amount of services and online information has grown rapidly, which brings challenges to provide personalized and trusted service recommendation for users with different preferences efficiently.2) To meet the requirement of complex applications, an optimal service composition plan should be generated for users. Traditional QoS-aware service optimization methods usually conduct service optimizations based on the QoS values provided by service providers. However, the QoS values provided by service providers could be affected by many factors such as dynamic network environment and commercial speculation, which makes the service composition plans not trustworthy.3) In dynamic and scalable cloud environment, the contract between the services and their users may be violated during the execution because of some uncertain reasons such as flash crowd, service failure and so on. Then the user’s application may not be completed as expected.In view of these challenges, we propose our solutions for trusted service recommendation, trusted service composition and trusted execution in cloud computing environment. Specifically, the main contributions of our work are listed as follows.1) To construct a trusted service application in cloud environment, a trusted service composition framework in cloud environment is proposed. The framework consists of four levels, which are service resource level, trusted service recommendation level, trusted service composition level and trusted service execution level. The service resource level is responsible for collecting resources such as services and user feedback from different cloud platforms through distributed service brokers. The trusted service recommendation level is to find the trusted previous users who have similar preferences with the target user, then recommend trusted services according to the feedback of these trusted previous users. When the user’s application requirements are too complex and cannot be implemented by single service, the trusted service composition level would divide user’s application requirements into sub-tasks, and then provide trusted service composition plans by selecting suitable service for each sub-task. Finally, in trusted service execution level, to deal with the uncertainties caused by the dynamics of the cloud environment, dynamic scheduling is conducted during the execution to optimize the global service scheduling and ensure stable execution of users’ application.2) To provide trusted and personalized service recommendation for users efficiently in the cloud environment, a keyword-aware personalized service recommendation method is proposed. Specifically, to obtain accurate preferences of users, we construct a preference-QoS keyword set and domain thesaurus. Based on the two data structures, the preferences of previous users can be extracted from their reviews for services. For different situations, two similarity computing methods are presented. According to the similarity between users, we can find trusted previous users for the active user. Then based on the ratings of trusted previous users, we can predict ratings of candidate services for the active user and recommend trusted services to him/her. Moreover, to improve the scalability and efficiency of the method, we implement it on a MapReduce framework in Hadoop platform.3) To improve the credibility of service composition plans in cloud environment, a QoS-aware trusted service optimization method based on history records and clustering is proposed. In cloud environment, due to network dynamics and commercial speculation, the QoS values provided by service providers may be inconsistent from actual QoS values. To address this issue, we take QoS history records to evaluate the utility of service composition plans. The authenticity of QoS history records can guarantee the service composition plans trustworthy. Meanwhile, to reduce the high computation complexity caused by large scale of service history records, we take a fuzzy-hierarchical clustering technique to cluster the QoS history records for each service. And then use the centroids of the subclusters to generate centroid-based composition plans. Based on the centroid-based composition plans, we can get the average utility scores of the corresponding service composition plans. Finally, the service composition plan with the highest utility score can be seen as the optimal and trusted service composition plan.4) Due to the dynamics of the cloud environment, uncertain events such as service failure, performance degradation may occur during the execution of applications. To guarantee the reliable execution of applications, an uncertainty-aware dynamic service scheduling method is proposed. Specifically, a trusted static service scheduling method is firstly proposed based on a reverse-auction-based service provisioning mechanism and the partial critical path strategy. To deal with the uncertainties during the execution, an uncertainty-aware dynamic service scheduling method is proposed. An uncertain model with four uncertain events is considered. To deal with uncertain events during the execution, intermediate workflows are created for dynamic scheduling to get a global optimal schedule, so as to support trusted execution for users’ applications.