A Knowledge Graph Driven Approach for Edge Analytics
3) Deployment Modularity: support for
multiple operating systems (e.g.,
Windows and Linux)
4) Messaging Support: support data
persistence, replication, and querying
(with support of time series databases
e.g., Cassandra, as well as document
stores, e.g., MongoDB®)
5) Connectivity Robustness: robust to
signal noise and data connectivity
limitations
6) Heterogeneous
Hardware
(HW)
Support: support multiple device
versions and deployments
and mobility in field deployments and
maintenance.
Large Industrial Construction and Mining
Manufacturing Company
The second use case is that of a large
construction equipment manufacturer,
building earth moving equipment, mining
vehicles, etc., and has the need of
interconnected and mobile sensors in large
open areas. The client has been deploying
machinery
capable
of
sending
communication and status feeds but in an ad
hoc manner with many iterations on device
firmware. To complicate the situation the
environment in which deployment resides is
prone to high mobility, equipment
breakdown from heavy usage and noisy
signal environment. As a result, the client
works within a high-maintenance/high-
touch environment prone to equipment
failure where vehicles need constant
attention. The client has been enabling
equipment over time but is tasked with an
extremely heterogeneous environment
where the approach is of "if it isn't already
broken, then don't fix it." While construction
and mining equipment has been updated
and maintained at a priority as it impacts the
core business model, the needs of sensors
and IoT devices is deprioritized in favor of
reliability and stability of signal data.
One of the more crucial use case
requirements is the robustness regarding
adverse site location issues such as lack of
connectivity. This client has several locations
where broadband or steady network
connectivity is a challenge and data
connectivity must fall back upon cellular or
satellite coverage at significant costs. That is,
high bandwidth streams may be cost
prohibitive to maintain beyond a few
Megabits/second. Therefore, data that is
originally of high value must be filtered and
down-sampled to lower rates when possible
and high-fidelity data streams must be
offloaded to physical storage and
transported (e.g., via helicopter) from the
remote locations. Thus, some analytics are
not streamed dynamically and require post-
processing.
Challenges
Additionally, the client currently has
ongoing requirements that the edge
framework must support; a wide range of
devices including laptops, edge gateways
and mini server racks. These requirements
are for providing three tiers of compute
capacity (processing, memory and storage)
IIC Journal of Innovation
As in the previous use case description, this
is an existing environment. This means that
the overlay of any solution must be
extensible to changing needs and support
current operations. Furthermore, in this
specific case the challenge is for a solution to
be highly resilient and reduce the need for
- 47 -