A Knowledge Graph Driven Approach for Edge Analytics
Messaging Backbone and Application Services
The messaging backbone( shown in Figure 7) allows for applications and data streams to independently interact with each other without having to previously determine which applications and sensor streams are present on each device.
We utilize NSQ 7 as the default message broker because it is fault-tolerant, dynamically scalable, has simplified configuration, no single points of failure and is primarily in-memory( bounded volatile memory with disk overflow) so that the applications have a lower latency to access sensor data.
Integral to the operations of the messaging backbone are the protocol converters which support it. These translators are dynamically instantiated to convert device specific data protocols( MQTT, Serial, etc.) to the consuming NSQ format applications. These converters or containers reside on the edge appliance. They are automatically created by the orchestration layer when the knowledge graph reflects that a new device has been onboarded and connected to the platform.
The application services component provides a software development kit( SDK)
for the application engineer to parameterize components of the application which vary across device instances. For example, the sensor stream topic names, IP addresses of the message brokers, etc. will all vary. The application engineer needs only to define a parameterized query and the application services will fill in the specifics( e. g., an accelerometer stream from the right front wheel well of a vehicle).
Figure 7: Edge Appliance Architecture
Containerization
Containerization, providing near native runtime specs 8, allows for the application engineer to encapsulate nearly all software dependencies of the target device. This allows the platform to orchestrate the
7
“ A realtime distributed messaging platform”, 2017. [ Online ]. Available: http:// nsq. io /
8
W. Felter and F. Alexandre, " An updated performance comparison of virtual machines and Linux containers," in IEEE Performance Analysis of Systems and Software( ISPASS), 2015.
IIC Journal of Innovation- 53-