Finding a way through the Fog from the Edge to the Cloud

Project description

The Internet of Things (IoT) promises to open up new applications that stand to benefit from numerous sensing devices in different environments. These applications may involve travel planning or congestion management in smart cities, proactive energy usage or health monitoring in homes, or coordinated emergency response.

The Internet of Things may be realized in practice, in relation to other computer infrastructure, as consisting of edge (sensing) devices, fog (intermediate computation and networking devices) and the cloud (elastic, but potentially too remote from edge devices to support timely decision-making). As a result, IoT architectures involve heterogeneous devices with widely differing capabilities, potentially supporting a diverse collection of applications with very different requirements. Furthermore, the objectives of application users (that relate to the specific needs of their application) may be quite different from those of the infrastructure providers (that relate to revenue and maximizing resource usage).

In this complicated environment, there are also a diversity of tasks to support, which include: infrastructure provisioning - deciding what devices to procure, and where to put them; resource allocation - deciding how to assign different edge/fog/cloud resources to a recently arrived application; and adaptation - how to revise resource allocation decisions in the light of changing demands or resource availability. In practice, all these decisions have to be taken in ways that reflect both application user and infrastructure provider priorities and constraints and their trade-offs. The latter suggests that no universally good solutions are expected: instead the suitability of a range of feasible solutions depends on how priorities and constraints are weighed.

This project is to investigate techniques that support all of provisioning, allocation and/or adaptation within a consistent framework; although these tasks are different, they share infrastructure features, application characteristics and user preferences. As such, in this project, the challenge is to: (i) model the infrastructure and applications in ways that support resource provisioning, allocation and adaptation tasks; (ii) identify ways of capturing the preferences and constraints of application users and infrastructure providers; and (iii) developing algorithms within a common framework for resource provisioning, allocation and adaptation over the model at (i) taking into account the preferences at (ii). This is potentially a very significant undertaking, so work will necessarily proceed incrementally, starting with simple models and strong assumptions, and incrementally relaxing the assumptions to move towards more comprehensive proposals.