A Knowledge Graph Driven Approach for Edge Analytics
applications 1 . As IoT deployments become
more numerous, scalable and easily
maintainable, architectural solutions are
paramount for meeting and sustaining the
demand of large, expanding, and elastic
device networks.
intimate domain knowledge and site
information details, namely: device expert,
application expert and the field engineer.
Collectively these engineers hold the domain
knowledge required to start and fluidly
maintain the network of edge devices at any
given location.
Current IoT deployments are extensions of
the edge-to-cloud paradigm where control
and analysis of signal data from edge devices
is centralized with elastic compute in mind.
While this methodology is the simplest in
terms of initial deployment and scalable
from a compute perspective, cloud
management does not imply scalable
maintainability with respect to a
heterogeneous device environment with
device types where there is a lack of protocol
or implementation rigor. While vertically
integrated solutions exist 2 3 that aid in this
type of deployment, they require a complete
migration of existing applications into the
new ecosystem (e.g., hardware, OS,
language, cloud tools, etc.). Additionally, the
edge-to-cloud paradigm assumes stable, low
latency and high bandwidth backhaul (uplink
and downlink) connectivity. However, IoT
deployments in remote locations may suffer
from limited backhaul connectivity and thus
render the timely processing and reaction to
data impractical.
As infrastructure communication is
orchestrated and edge devices are brought
online there is subtle knowledge of devices
that must be documented to onboard new
and similar devices. IoT applications are
moving beyond simple sensor data storage
and towards analytics applications where
novel architectures are required to enable
the horizontal scalability of device additions,
reduce instantiation time, boot strap
domain knowledge, enable system reuse
and migrate compute and intelligence
toward the edge. When moving from
domain to domain, isolation of knowledge
kept with the domain experts becomes a
limiting factor of rapid deployment when
working with a mature portfolio of edge
framework
configurations
(varying
hardware, varying device driver versions,
language
attributes,
communication
protocols, etc.).
To avoid refactoring, increased man-hours
and to promote rapid onboarding and
enable knowledge retention, we developed
a solution that ensures that the instantiation
Additionally, knowledge of the edge
networks relies on key personnel with
1
L. Columbus, "Internet Of Things Market To Reach $267B By 2020," Forbes, January 2017. [Online]. Available:
http://www.forbes.com/sites/louiscolumbus/2017/01/29/internet-of-things-market-to-reach-267b-by-2020/#2343106e609b.
2 “GE Predix IoT Platform," 2017. [Online]. Available: https://www.ge.com/digital/predix
3 "Thingworkx IoT Platform," 2017. [Online]. Available: https://www.thingworx.com/
- 44 -
September 2017