rafehi
commented
5 years ago Google makes a distinction between three types of device data:
- Metadata
- State
- Telemetry
They also define commands (instructions to devices to perform actions) and operation data (e.g. cpu, battery level etc, which can be stored as either state or telemetry data).
Modelling of assets is critical to the system and dictates whether we can compare fields across different asset types etc.
There are a number of different approaches we can take. Schemas allow us to know what type of data an asset expects. Asset hierarchies allow us group assets together. Fieldsets can be used to group fields that are assignable to an asset. Asset inheritance allows us to create more specific types of assets.
Examples: Schemas. Schemas allow us to know the type of an asset's field. For example, a car might have a
current_speedfield. The schema might tell us that it measures speed, in units km/hr. This then allows us to convert that field to other units, compare it against other vehicle types, etc. Asset hierarchy. A wind turbine is composed of blades, cooling unit and gearbox. Fieldsets. All assets might be expected to have a name, manufacturer, owner adnd status. These fields could be grouped into a default fieldset, which is then assigned to the asset type. Similarly, for vehicular assets, we might find that they all havedistance_travelled,current_speedandcurrent_location. We can create a mobile fieldset and assign that to the car asset type. This allows us to easily compose new asset types for known fields. Asset inheritance. It might be desirable to inherit more specific asset types from less specific types. For example, we might have anelectric vehicleasset, which extends avehicle, and adds fields such asbattery_capacity.