Riccardo Lancellotti

The Fog computing paradigm makes use of dispersed, diverse, and resource-limited devices located at the network edge to effectively implement Internet of Things (IoT) application services that demand low latency and substantial bandwidth. At the same time, the adoption of microservice-based architectures in the IoT domain is on the rise due to their ability to align with the swift evolution and deployment demands of highly dynamic IoT applications and to elastically scale to fulfill load demands. In complex environments like Fog federations, characterized by highly heterogeneous computing and networking resources, the effective allocation of microservices to available nodes, while ensuring compliance with required Quality of Service (QoS) constraints, represents a significant challenge. In this paper, we present the design and implementation of OptiFog, a comprehensive framework that enables users to model, simulate, and validate microservice placement solutions within a realistic testbed environment. Compared to state-of-the-art approaches, OptiFog offers developers a controlled environment for experimenting with placement solutions while providing the assurance that the resulting deployments will meet the targeted QoS requirements in real-world scenarios, specifically in terms of service execution time and energy consumption of Fog nodes. To demonstrate the feasibility of the proposed approach, we implemented and evaluated a representative use case, involving both sub-optimal and optimal microservice placement, and utilizing real-world microservices drawn from the IoT domain.

The evolution and widespread adoption of virtualization, service-oriented architectures, autonomic computing, and utility computing have converged, giving rise to Cloud Computing, which enables the on-demand delivery of software, hardware, and data as services. As Cloud-based services become more numerous and dynamic, developing efficient service provisioning policies is increasingly challenging due to the impact of estimation errors in the operating parameters of applications such as execution time or request rate for each application and for each service. In this paper, we take the perspective of Software as a Service (SaaS) providers which host their applications at an Infrastructure as a Service (IaaS) provider. Each SaaS needs to comply with quality of service requirements, specified in Service Level Agreement (SLA) contracts with the end-users, which determine the revenues and penalties based on the achieved performance level. SaaS providers should maximize their revenue and minimize their cost by deciding how many VMs for each class (reserved, on-demand or spot) are to be used, while IaaS must decide the price of VMs in order to maximize his revenue. The problem is modeled using a Generalized Nash Equilibrium Problem (GNEP). The uncertainty in the application parameters, that are not known until run time, is specifically addressed in this paper using a cardinality-constrained robust optimization approach to formulate a robust version of the SaaS problem. Furthermore, we propose a best-reply algorithm to identify a robust GNE, ensuring that the solution remains feasible even under the worst-case scenarios of uncertain parameters. An analytical model is developed to derive the equilibrium. For validation, the analytical model results are compared with simulations that reflect real situations, characterized by more realistic parameter distributions and limited buffer capacities.

The widespread adoption of Artificial Intelligence applications to analyze data generated by Internet of Things sensors leads to the development of the edge computing paradigm. Deploying applications at the periphery of the network effectively addresses cost and latency concerns associated with cloud computing. However, it generates a highly distributed environment with heterogeneous devices, opening the challenges of how to select resources and place application components. Starting from a state-of-the-art design-time tool, we present in this paper a novel framework based on Reinforcement Learning, named FIGARO (reinForcement learnInG mAnagement acRoss the computing cOntinuum). It handles the runtime adaptation of a computing continuum environment, dealing with the variability of the incoming load and service times. To reduce the training time, we exploit the design-time knowledge, achieving a significant reduction in the violations of the response time constraint.

In the last few years, fog computing has been recognized as a promising approach to support modern IoT applications based on microservices. The main characteristic of this application involve the presence of geographically distributed sensors or mobile end users acting as sources of data. Relying on a cloud computing approach may not represent the most suitable solution in these scenario due to the non-negligible latency between data sources and distant cloud data centers, which may represent an issue in cases involving real-time and latency-sensitive IoT applications. Placing certain tasks, such as preprocessing or data aggregation, in a layer of fog nodes close to sensors or end users may help to decrease the response time of IoT applications as well as the traffic towards the cloud data centers. However, the fog scenario is characterized by a much more complex and heterogeneous infrastructure compared to a cloud data center, where the computing nodes and the inter-node connecting are more homogeneous. As a consequence, the the problem of efficiently placing microservices over distributed fog nodes requires novel and efficient solutions. In this paper, we address this issue by proposing and comparing different heuristics for placing the application microservices over the nodes of a fog infrastructure. We test the performance of the proposed heuristics and their ability to minimize application response times and satisfy the Service Level Agreement across a wide set of operating conditions in order to understand which approach is performs the best depending on the IoT application scenario.