Dynamic service adaptation for cost minimization of composite cloud service under stochastic runtime performance

• Main Authors: Lu,Sin-Ji, 盧信吉
• Other Authors: Huang,Kuo‑Chan
• Format: Others
• Language: zh-TW
• Published: 2016
• Online Access: http://ndltd.ncl.edu.tw/handle/4j76rp

碩士 === 國立臺中教育大學 === 資訊工程學系 === 104 === In recent years, cloud computing has become the most popular next-generation computing platform. Software as a Service (SaaS) based on Service-Oriented Architecture (SOA) is transforming how developers construct software and how users access software service on cloud computing platform. Service selection is one of the challenging problems when developing composite cloud services. Service selection can be seen as a optimization problem under the constraint of the service level agreement (SLA) between service providers and users. This thesis explores the service selection problem in dynamic cloud environments where service performance varies dynamically. The goal is to select a set of appropriate services to reduce the total cost of running a composite cloud service as most as possible, considering the constraint of SLA. This thesis presents three new service selection methods. The first method is an iterative approach based on Chebyshev’s theorem and integer linear programming. The second method adopts a dynamic service adjustment strategy during the execution of a composite cloud service. The finish of each task will trigger service reselection for remaining tasks based on the updated constraint. The third method applies nonlinear programming. It takes dynamic performance variation into consideration in the target function of the formulated nonlinear programming problem, avoiding the iterative computing process in the previous two methods. We conducted a series of simulation experiments to evaluate and compare the three service selection methods and compared them to the previous approach in the literatutre. Experimental results show that our methods outperform the previous approach, having potential to further reduce the total cost of running composite cloud services.