Auto-scaling Strategy for Amazon Web Services in Cloud Computing

摘要

Auto scaling mechanisms have become a typical paradigm in cloud computing environments. Such mechanisms can increase or minimize the number of virtual machines according to user demands, consequently achieving pay-per-use objectives. However, auto scaling mechanisms provided by infrastructure-as-a-service providers must strictly follow user-defined thresholds, the drawback of such mechanisms is that they cannot respond to real-time Internet traffic loads by following user-defined thresholds. Therefore, we propose a dynamic threshold adjustment strategy that can expedite the creation of virtual machines according to workload demands. The proposed strategy can reduce the web application response time and error rate when the system is under a heavy workload. In addition, it can expedite the release of virtual machines to reduce virtual machine running time when the system is under a light workload. According to our experimental results, we found that CPU-intensive web applications require an excellent threshold control strategy. Therefore, the proposed strategy can satisfy this requirement by effectively reducing the response time of applications, virtual machine running time, and error rate.