BLE MESH with variable number of Sensors

[dbuijsma]

We have an application where we have a large number of sensors, some relay nodes and one or more gateways. The sensors and relay nodes are provisioned by the gateways and then start publishing data. Each gateway has a number of sensors publishing data to its Unicast address. That number of sensors is unknown in advance. 

As I understand it, for each Sensor Server Model, I need a Sensor Client Model instance on the Gateway. I've been looking at how to implement this in the most practical manner and the issue I am running into is that we don't know what the number of sensors is in advance. I could solve that in several ways:

1) Let all Sensors subscribe to the same Sensor Client and implement my own code on top of that to deal with de-multiplexing the sensor data. That would not be according to the BLE MESH spec, but consumes fewer resources.

2) Pre-allocate more Sensor Client Models than I'll ever need (wasteful)

3) Dynamically allocate Client Models: according to the BLE Spec that means I need to re-provision the Gateway every time a new sensor joins (very wasteful)

The same is true for the Health Server in each of our Nodes. Do I have to create an instance for every single one? Is there a better way? 

Thanks!

[Bjørn Kvaale]

You only need to have one client model on the gateway, which can talk to multiple server models. This is what happens with the generic on off client & server models, as well as the health model.

By setting the publish & subscription addresses correctly, one client model should be able to speak to multiple server models & vice versa.

What is the maximum number of sensor nodes we are talking about here? For the gateway, do you want to use a 52840 device or do you want to provision via interactive pyaci?

[dbuijsma]

Thanks Bjorn! We use the interactive pyaci with the nrf52832.