A Knowledge Graph Driven Approach for Edge Analytics
deployment of applications across any
device if it supports Docker containerization,
an extensively used industry standard
technology in the cloud which eases the
skills transition for the application engineer.
Our platform provides an interaction layer
on top of the containers for setting
configurable runtime parameters. For
example, we can declare that some
applications require access to specialized
hardware such as a GPU. After querying the
knowledge graph, the orchestrator layer
enables launching of the container on the
specified device target while exposing the
required hardware components via Docker
runtime parameters.
U SE -C ASE I MPLEMENTATION
We can leverage a combination of these key
solution-architecture
components
(knowledge graph, messaging backbone and
containerization) to address the various
challenges we have previously identified.
While all components are used in each use
case, highlighted are the key contributions of
our framework to resolving some of the
major requirements for each use case.
Large Oil and Gas (O&G) Services Company
In the case of the O&G service provider the
most impactful contribution to addressing
the key requirements and important
constructs from our edge framework for this
client is as follows:
applications on the edge devices. The
knowledge-graph contains the metadata
about each application’s software
dependencies
and
hardware
requirements. It also contains metadata
about the capabilities provided by each
edge device instance and the various
sensors attached to it from which the
client can query and infer all the edge
devices where an application could be
deployed. The graph provides querying
and inferencing capabilities of new and
old devices alike along with their
applications by capturing associated
capability metadata. Furthermore, the
knowledge graph promotes a structured
environment in which onboarding
heterogeneous
edge-hardware
is
simplified for technology roadmaps and
migration plans.
Containerization - Heterogeneous SW
Support/Deployment
Modularity:
Applying containerization decouples the
edge-applications from one another and
provides an abstraction over the
underlying sensor hardware. When
combined with the application-metadata
captured in the knowledge-graph, it
provides the capability to encapsulate,
isolate dependencies and utilize a wide-
variety of available CEP and database
products or AI frameworks (proprietary
or open source) without worry of
conflicts between to co-located
application containers.
Large Industrial Construction and Mining
Manufacturing Company
Knowledge graph – Extensibility and
Agnostic Support/Heterogeneous HW
Support: The knowledge graph enables
the client to extend support for any
vendor or 3rd party provided
Similar to that of the O&G use case, the edge
framework components address all the key
challenges and requirements in the
- 54 -
September 2017