HERMES: Federating Fog and Cloud Domains to Support Query Evaluations in Continuous Sensing Environments

摘要

As networked sensing devices continue to proliferate, the storage and processing capabilities of fog or edge devices have also increased substantially while offering low energy consumption and hardware costs. These complimentary technologies enable unique opportunities for performing decentralized analysis on the edges of the network while also leveraging the capabilities of public and private clouds for coordination and long-term storage. Our framework, HERMES, enables federated query evaluations that transition between cloud and fog nodes seamlessly. The system selectively samples from observational streams to reduce communication and memory consumption on fog nodes; our evaluation over a real-world observational dataset demonstrates an 89% reduction in dataset size while maintaining a mean absolute error of less than 0.25%.