Since the Edge compute device can have device attributes, when should I use those compared to attributes defined on a peripheral device?
Lets says I have a plc that the edge talks to using modbus/tcp. a few sensors are hooked up to that PLC.
Couldn’t I either define the attributes on the edge, or on the peripheral ?
If so, what are the benefits with doing it one way vs the other?
Hi
For us we have stuck to representing a PLC as a peripheral and keep the agent to represent all activity specific to the agent. (ie its own temp, memory, storage)
One good reason is it allows easier differentiation between no activity from peripheral vs no activity from the agent.
Other reasons, - we often have an Edge Agent polling via modbus multiple MODBUS devices.
Again makes it easier to keep workflows and ttributes separate from each other.
Thanks for that Info Tim.
May I ask which edge compute device you use most and are happy with?
HI Lars
We have pretty much used raspberry pi’s 
Some have been out in the field now for over 3 years. (Went out before we started using Losant.
We use a UPS board on them. They are in steel cabinets and experience 40+ deg C outside temp. Cabinets get way hotter
The only ones that have failed have been due to
We use the Edge Agent and our own in house developed agent (started that in 2015) which is python based on talks to engine controllers via MODBUS. We still use it for a specific class of controller as it works out of the box parsing register definition files we download from the controllers. (The register addresses are not fixed in the controllers depending on what options they have).
Have you had the need for an edge device with 2 Ethernet ports?
One for Modbus Lan and one for WAN.
HI
Yes, I would like to get something like that. MODBUS is awfully insecure and I would prefer to run a firewall on the edge to restrict what/who can connect - (ie nothing
Have looked at some of the small openwrt based routers with LTE modem. As we also add a modem or other radios to get out of mining pits. But these devices all seem to be too small (memory/cpu) to run docker and the edge agent. Our agent on the other had runs fine in 128MB systems.
The PI isn’t the perfect fit, but reliably available, and consistent OS has been the main thing.
My ideal device would have 2 ethernet ports and an optional LTE modem (not USB based - had too many problems in the past not being able to reset them with out power cycling). Something like a Balena Fin https://www.balena.io/fin/ and just add the USB ethernet. it supports a Mini PCI express port so could add the LTE interface using that or an Ethernet. (I don’t need both at the same time).
Our requirements change each time we deploy. Custom built network, different modbus devices. Some with no MODBUS, etc… so each job is quite bespoke.
T
I have used the Balena cloud managed image on a regular pi3 and that’s working good. But back to 2 ports, this looks interesting https://revolution.kunbus.com/revpi-connect/