Multiple objectives dynamic VM placement for application service availability in cloud networks

Ensuring application service availability is a critical aspect of delivering quality cloud computing services. However, placing virtual machines (VMs) on computing servers to provision these services can present significant challenges, particularly in terms of meeting the requirements of application service providers. In this paper, we present a framework that addresses the NP-hard dynamic VM placement problem in order to optimize application availability in cloud computing paradigm. The problem is modeled as an integer nonlinear programming (INLP) optimization with multiple objectives and constraints. The framework comprises three major modules that use optimization methods and algorithms to determine the most effective VM placement strategy in cases of application deployment, failure, and scaling. Our primary goals are to minimize power consumption, resource waste, and server failures while also ensuring that application availability requirements are met. We compare our proposed heuristic VM placement solution with three related algorithms from the literature and find that it outperforms them in several key areas. Our solution is able to admit more applications, reduce power consumption, and increase CPU and RAM utilization of the servers. Moreover, we use a deep learning method that has high accuracy and low error loss to predict application task failures, allowing for proactive protection actions to reduce service outage. Overall, our framework provides a comprehensive solution by optimizing dynamic VM placement. Therefore, the framework can improve the quality of cloud computing services and enhance the experience for users.