Drive Your Application Through the Fog
by Domenico Siracusa, Daniele Santoro from the FogAtlas project of Fondazione Bruno Kessler
Cloud technologies have revolutionized how a large fraction of computer software is built, maintained, and used: a lot of services we interact with every day, such as e-commerce or mail servers, or even office software, are now run from remote data-centers in the cloud. Such an approach allows service providers to quickly provision and consumes, on a large scale, computing, networking, and storage resources from cloud providers, without owning significant infrastructure themselves. This, however, may come at the expense of the quality of the service provided. Consider the case of a factory deploying cameras and a video processing software to enforce social distancing: the factory could deploy the necessary computing infrastructure to process feeds locally, or pay for a lot of reliable bandwidth to a remote data-center (both expensive, and the latter may also add excessive delay). Ideally, they’d want their cloud provider to offer the ability to provision resources close to the factory, which they could simply use on an as-needed basis, to safeguard performance, reliability and perhaps honour privacy and data locality laws.
To address such and other scenarios, the industrial and academic communities elaborated the concept of fog computing, which attempts to make use of computational resources deployed near the edge of networks, close to consumers, where data is often generated and consumed while maintaining the cloud’s successful service interface. This preserves the reliability, scalability, and maintainability of the Cloud and adds efficient, localized processing preserving data locality and privacy.
This technology is being tested for production deployments in a variety of sectors. Most existing fog products are integrated hardware and software solutions to create a fog instance on-premise. Their primary focus appears to be creating local infrastructure, not managing services intelligently. This will become a problem as soon as several services, competing for precious edge resources, are deployed on the same infrastructure.
Our solution, FogAltas, fills this gap by offering cloud-like service management for distributed fog environments. It delivers automated management of applications between the cloud and edge while ensuring the quality of service and reliability. In brief, FogAltas offers four core functionalities: application modeling, service-aware orchestration, zero-touch provisioning, and fleet management. Thanks to its application composer, an operator can model the applications to deploy through a graphical user interface and define operational requirements in terms of computing, networking, location, etc. For example, for a smart-lock application, where a camera records people waiting to enter a location, a face-recognition module processes this feed to recognize the bystanders, and a directory service finally matches these identities with their permissions to enter the specific premises, a very stringent latency requirement may be needed between the camera and the service processing its video feed, while no such requirements would exist between the latter and a statistics collection module, which may as well run remotely in the cloud. Then, the FogAtlas service-aware orchestration algorithms take care of deploying the individual modules optimizing the utilization of the infrastructure while honouring the requirements of each application. The resulting recipes are then automatically handled by FogAtlas’ zero-touch provisioning module, meaning it manages deployments without requiring any further input from its operator. Finally, the deployed modules are constantly monitored to ensure the smooth running of every application, even as the load of the infrastructure changes over time. FogAtlas can also handle infrastructure churn (e.g. nodes deactivated for maintenance or replacement) and the non-disruptive deployment of application upgrades.
To summarise, FogAtlas delivers advanced service management in heterogeneous distributed computing environments, driving applications through the fog and enabling the so-called cloud-to-edge continuum. It is like the control tower of an airport or the workforce manager in a logistic company. Thanks to FogAtlas, infrastructure managers will be able to implement the fog computing digital change and make their IT processes smarter, smoother, and cheaper.
