Resources
-
Project
BRA2Cloud
This project investigates the application of agent-based architectures to create a resilient environment using unsecured transient servers to offer trusted services or run applications using Cloud Computing idle resources. Exploring idle resources is an efficient way to save energy and money (e.g., reuse unused CPU and memory to provide services and run applications). The BRA2Cloud architecture combines machine learning and a statistical model to predict instance survival time and helps to refine fault tolerance parameters to provide trusted services, reducing monetary cost. This model compiles and analyses Amazon EC2 Spot Instances’ historic price change data to predict revocation events. Our agents pursue an efficient usage of Spot Instances, providing a novel resilient environment between users and cloud resources, through machine learning, to predict revocation events and define suitable Fault Tolerance mechanisms with their respective parameters. This is a key step toward successful and efficient usage of these instances to provide trusted services with minimal interruptions at cheapest prices. Experiments indicate that this model can be used under realistic working conditions with better use of idle resources.