Research On The Problems Of Application Placement And Collaboration In Cloud Computing

With the rapid development and popularization of Internet, there are more and more web applications. These applications are becoming large and complex. Moreover, the collaboration of the applications is gradually becoming more popular. Therefore, the development trend of the applications is collaboration. To deal with this problem, how to make the applications flexible and efficient collaboration is currently a hot issue. The rapid development of cloud computing technology provides a good chance to this problem. The cloud computing technology can provide high availability and scalability. However, how to build a collaboration application based the cloud flexibly, rapidly and automatically is still a big challenge for the current.As an important part of cloud computing, the application placement and collaboration problem in the PaaS has become a hot issue of current research. This paper is dedicated to improve the application quality of service, and reduce the cost of the cloud service provider in the application placement process and automatic build the collaboration applications. However, due to the characteristics of cloud applications, highly dynamic and autonomous, load changes rapidly, there are several problems need to be solved:(1) The application placement and adjustment. The application placement in cloud platform has a great impact to the cost of cloud platform and the efficiency of the application and collaboration. So it is a challenge to satisfy all application efficiency and reduce the cost of application collaboration while minimizing the cloud nodes occupancy to reduce the cost of cloud service provider. (2) The automatic building of collaboration meta-model. How to use the available application in the cloud, to automatically build collaboration meta-model to meet the individual needs of users is the key technology and research hotspot in cloud computing and a major challenge. (3) The self-evolving of collaboration meta-model. How to efficiently and automatically evolve the collaboration meta-model according to the information of collaboration application instantiation, to make the collaboration meta-model closer to the needs of users and reduce the complexity of the collaboration application instantiation is one of the core issues of the application collaboration.This thesis aims at the application placement and adjustment, the automatically building and evolving of collaboration meta-model. The main research works and contributions are as follows:(1) Due to the contradiction of improving the application performance and reducing the cost, a performance-cost balanced application placement strategy is proposed. The strategy can balance these two conflicting objectives to provide better quality of service while reducing the cost.To make the applications running efficiently, reduce the cost of application collaboration and minimize the cloud node occupied in the application placement process, it needs to resolve the cloud application placement problem. To solve this issue, the application placement problem is modeled as an application placement diagram. Then the Pareto optimal theory is used to propose a performance-cost balanced application placement strategy in this paper. The strategy can balance the two conflict goals, performance and cost, effectively. The experimental results show that this method can solve the application placement problem and find a better placement solution.(2) Due to the dynamic change of applications in cloud platform, an adjustment method of application placement is proposed. The application placement can be adjusted efficiently and improve the application collaboration performance at the same time by using this method.To adjust the application placement timely when applications in the cloud platform changes dynamically, ensure the efficiency of all applications in platform and reduce the cost of application collaboration and platform, it needs to adjust the application placement in cloud. To solve this issue, this paper proposes a collaboration application placement adjustment method. The method models applications and collaboration costs between applications as a diagram. Then according to the existing application placement and the collaboration cost, it can find a placement adjustment solution rapidly when the statuses of applications are changed. The experimental results show that this method has good performance and strong scalability. It can improve the application collaboration performance efficiently.(3) Due to the automatically building the collaboration meta-model from a large number of applications, a collaboration meta-model building method is proposed based on extended planning graph. The method can reduce the search space in the process of building collaboration meta-model and quickly find multiple collaboration meta-models that satisfy user needs.To make the applications in platform can work collaboratively and provide multiple collaboration applications for the users, it needs to auto build collaboration meta-model. To solve this issue, this paper proposes a method to build collaboration meta-model based on extended planning graph. The method by extending the traditional planning graph can build multiple collaboration meta-models by search once. In addition, it improves the efficiency of building collaboration meta-model. The experimental results show that this method has good performance and strong scalability, can be applied to build collaboration meta-model when the candidate application set has large-scale applications.(4) Due to the automatic evolution of collaboration meta-model, this paper proposes a self-evolving method of collaboration meta-model. The evolved collaboration meta-model is closer to the user needs. It can improve the collaboration application instantiation efficiency and reduces the complexity of using collaboration applications to users.To improve the efficiency of collaboration application instantiation and reduce the complexity of using collaboration application, it needs to study the self-evolving problem of collaboration meta-model. To solve this issue, a self-evolving method of collaboration meta-model is proposed in this paper. The method models collaboration application as points in multidimensional space and evolves the collaboration meta-model to multiple collaboration meta-models that meet different demand by using clustering algorithm. The experimental results show that this method has good performance and can significantly reduce the distance between collaboration application and collaboration meta-model. Moreover, it can reduce the complexity of using collaboration applications to users and the collaboration application instantiation time.