Intelligent Auto Scaling and Profit Optimization in Cloud Computing

Abstract: Cloud computing offers resources as and when needed but poses a great challenge in terms of efficient management, especially when the resource demands are highly variable. Literature has explored various approaches in this regard, including resource allocation, auto-scaling, virtualization, workflow scheduling, cost optimization, and QoS management. Predictive, adaptive, and reinforcement-learning based auto-scaling techniques are some of the emerging trends in this area, which address the shortcomings of traditional CPU or memory-threshold-based auto-scaling methods. While traditional approaches are simple to implement, they are often associated with the risk of over-provisioning in case of low load and poor responsiveness to sudden traffic surges. Some recent works have focused on addressing these limitations, as well as exploring horizontal and vertical scaling, serverless platforms, ML-centric workloads, and hybrid and multi-cloud environments, among others. Taken together, these approaches aim to improve performance, utilization, latency, cost-effectiveness, scalability, and SLA compliance, among others. However, the issue of provider profit maximization still remains under-addressed in this context, with most existing contributions focusing on a limited set of optimization criteria. This work makes an attempt to address this challenge by considering the interplay between intelligent auto-scaling, efficient resource allocation, and profit maximization in cloud environments.

Keywords: Predictive Analysis, Dynamic resource Provisioning, Load Balancing, Cost Optimization, Workload Prediction, Resource Optimization.