Edge Intelligence: The Central Cloud is Dead – Long Live the Edge Cloud!
Visual analytics – to explore visually IoT
data stored on IoT gateways. IoT data
analysts can visually inspect the data
collected at the edge. For example, after
an alert has been sent to the cloud, an
analyst can dig into the details which led
to the alert.
3rd party application hosting – to allow
3rd party application containers to be
run on edge hardware, allowing
decoupling between hardware and
applications. For example an edge
gateway might be used to run several
services (camera, access control, AC
management, elevators)
End-to-end sophisticated management
system
–apply
software-defined
networking and other paradigms from
5G and other sources, to enable new
business models on the edge intelligence
based on tightly integrated services and
networking.
C ONCLUSIONS
The following conclusions can be drawn
from the review and analysis undertaken in
this article. The potential of edge
intelligence in the Internet of Things,
requirements, current technology gaps, and
standardization need to be addressed to
realize that potential:
Edge intelligence (EI) is edge computing
with machine learning capabilities. Data
can be analyzed and decisions can be
made by algorithms at the edge, i.e. very
close to where the data is collected and
where the machine and other equipment
is controlled. This makes it possible to
react
autonomously
(without
a
IIC Journal of Innovation
- 29 -
connection to the cloud) and with very
short response times.
Containerization will be important to
deploy and manage edge intelligence
consistently
and
economically.
Containerization allows to encapsulate
functionality, for example a machine
learning algorithm, in a software
package which can be deployed
anywhere, e.g. in a public cloud, a private
cloud, on premise, a micro data center
on a shop floor, in a vehicle, within a 5G
network or on an IoT gateway. This
increases the efficiency of implementing
new software development and allows
optimizing the deployment according to
customer specific requirements without
additional programming effort. There
currently exist no standards directly
covering this technology, although there
are many open-source initiatives, such as
Docker and OCI.
Common data models for edge
computing node communication are
essential to the success of edge
intelligence. A common data model
enables the interoperability between
devices, communication protocols and
software solutions from different
vendors.
Micro data centers will become more
important in this process, for a number
of reasons, including providi