A learning-based cost management system for cloud computing

Cloud cost management is an emerging and challenging issue as increasing number of business are moving their on-demand computation workloads to the public cloud. Although the business model of “pay-as-you-use” is used by many public cloud providers, the customers still usually pay much more than what they are actually using. To better identify the cost optimization opportunities, in this paper, we introduce four metrics to measure the efficiency of provisioned cloud computing resources and services: resource utilization, instance utilization, cost utilization, and cost saving efficiency, and apply temporal learning algorithms to predict their future values and to detect anomalies. We design a cost management system which can automatically monitor the cloud environment and track the changes of these four efficiency in real time, based on which actionable advisories are provided to the cloud consumers for their cloud cost and performance optimization. Using the Amazon Web Service Cloud as the public cloud provider, our designed system can effectively manage the cloud environment that automatically fit the demands of cloud computing applications.