The conventional cloud-centric IoT application does not meet the requirement of the quick reaction of time-critical applications. The idea of edge and fog computing arrived to overcome the limitation of cloud computing in the domain of IoT applications. In fog computing, the workloads get distributed across the fog devices, and the placement of services in the nearest fog devices drastically reduces the network delay, connectivity, and reliability issues and delivers real-time capabilities as an extension, it reduces energy consumption and network overhead in the case of large sensor networks. However, achieving seamless interoperability, platform independence, and automatic deployment of services becomes the major challenge over heterogeneous fog devices. This work proposes an adaptation of the OASIS - Topology and Orchestration Specification for Cloud Applications (TOSCA) for service deployment for fog computing. With TOSCA, we build an integrated and standards-based fog computing federation framework that abstracts all the heterogeneity and complexities and offers a user-friendly paradigm to model and dynamically deploy fog services, on-demand, on the fly, from a remote system. The framework uses (i) standard TOSCA Service Template for application modeling. (ii) Docker Containers (OS-level virtualization) for platform independence, and (iii) Docker Swarm (container orchestration tool) manages seamless coordination and cooperation across heterogeneous fog devices. Moreover, the fog application is dynamically deployed by a lightweight TOSCA-compliant orchestrator on a set of out-of-the-box fog devices. The case study on the framework indicates convincing resource consumption by the orchestration process on these resource-constrained fog devices. The framework realizes various fog applications through the dynamic deployment of custom or user-developed services on the fly.

Thesis: PDF link