Discrete-Event Simulation of the Impact of Spot Instance Time Prediction in IaaS Cloud Computing

The demand for scalable and cost-effective computing resources has driven the widespread adoption of Infrastructure as a Service (IaaS) in cloud computing. Spot Instances, a key offering in IaaS, provide customers with access to unused cloud capacity at significantly reduced costs, though with the risk of sudden interruptions. These interruptions make it hard for the customer to estimate the time it takes for her job to finish. In this study, we conduct a discrete-event system simulation to show the impact of helping customers estimate their time in the system more accurately. The results suggest that using machine learning techniques to predict the time customers must spend in the system can be beneficial for both customers and the provider.

Keywords:

Spot Instances , Infrastructure-as-a-Service (IaaS) , Cloud Computing , Discrete-Event Simulation , Machine learning , Prediction