IEEE Consumer Electronics Magazine - May/June 2023 - 54

Edge Computing
TABLE 1. Summary of Exiting Edge Computing Solutions.
Edge
Cloud
Solution
KubeEdge
K3s
AWS
Outpost
OpenHorizon
EdgeX
Foundry
Fledge
Managed
Yes
No
Yes
Yes
Yes
Yes
Edge
only
Focus
Possible Containers
Yes Containers
VMs
No
Containers
No Containers
No
No
IoT
I4.0
in accordance with the Cloud computing principles.9
They provide, respectively, an Infrastructure-as-a-Service
(IaaS) and a Platform-as-aService
(PaaS) over a shared virtualized infrastructure
residing in large centralized data-centers
and they both consider a logically
centralized controller. The controller comprises
multiple physical instances of the same software
components to address reliability and scalability
requirements. Due to the amount of real-time
stream-based monitoring information required
by controllers to manage the infrastructure,10 a
dedicated network for control and management
planes is normally provisioned. Moreover, servers
are segmented and segregated in clusters to
improve resiliency and operational management.
For what concerns Edge computing, multiple
solutions have appeared in themarket,mainly consisting
in an adaptation and extension of existing
Cloud solutions to the near-Edge. Such solutions
focus on bridging Edge resources under the control
of existing Clouds, such as KubeEdge, AWS
Outpost, or OpenHorizon, de facto extending the
Cloud to the Edge. Others instead have put the
focus on vertical markets: EdgeX Foundry focuses
on IoT applications and their requirements,
whereas Fledge specializes in Industry 4.0 and
automation. Still others focus on Edge-only deployments,
such as K3s, allowing the management of
Edge-only infrastructures without the need of
being connected to a Cloud. A brief summary of
existing Edge solutions is presented in Table 1.
Conversely, far-Edge is still a subject of
research and no production-ready solutions are
available yet. Bearing in mind that far-Edge
infrastructure is by nature decentralized, different
approaches are being proposed to address
such aspect. For instance, Ciavotta et al.11 analyze
the impact of the far-Edge on vertical markets
like Industry automation. Others, like Lebre
et al.12 and Xiong et al.13 propose to replace the
centralized components in nowadays Edge solutions
with distributed counterparts. Besides,
Kayal,14 Goethals et al.,15 and Munir et al.16 focus
on bridging cloud to support resource constrained
devices via ad-hoc allocation policies
and reduced resource utilization, with particular
attention on how to employ those devices in a
mixed scenario of near- and far-Edge.
To better compare Edge and Cloud architectures
and solutions, we select Kubernetes as a
reference open-source project, which presents
tailored versions for the Cloud (i.e., Kubernetes)
and the (near-)Edge (i.e., KubeEdge and K3s). It
is worth noticing that we selected these projects
because of the following reasons:
i) there are existing works on extending these
solutions for the far-Edge;
ii) the selected solutions provide a Cloud-like
model to application developers;
iii) the selected solutions present a good level of
maturity required for the experimental evaluation
performed in the section " Experimental
Evaluation; "
iv) the selected solution are widely used in
industry and academia to build Edge computing
testbeds.
Figure 2 illustrates the architectures for
Kubernetes, K3s, and KubeEdge. All these architectures
are composed of two different sets of
components: 1) control plane and 2) compute
plane. The components of the former provide
the management and orchestration functionalities
across the infrastructure. While the components
of the latter provide the deployment and
monitoring functionalities for the nodes composing
the infrastructure. A brief overview of each
solution is provided in the following.
Kubernetes: It is a portable, extensible, opensource
platform for managing containerized
workloads. It provides a PaaS offering:
i) service discovery and load-balancing;
54
IEEE Consumer Electronics Magazine
IEEE Consumer Electronics Magazine - May/June 2023

Table of Contents for the Digital Edition of IEEE Consumer Electronics Magazine - May/June 2023
