nRF91: Example results of RAI for samples/cellular/udp?

[JordanYates]

NCS has a sample showing sending UDP packets over a cellular network, with optional control over the RAI socket options:

https://nrfconnectdocs.nordicsemi.com/ncs/latest/nrf/samples/cellular/udp/README.html

Is there any documentation on how the sample behaves with the various options enabled on a network that support AS-RAI?

It would be very helpful to have examples on the RRC Active -> Inactive timings with the various combinations of RAI enabled as a reference point for the behavior of local operators.

For example on my local operator, it appears as though `NRF_RAI_LAST` and `NRF_RAI_NO_DATA` are behaving the same way, with the modem transitioning back to idle after ~5 seconds after a `zsock_send`, compared against a PSM Active Time of 16 seconds.

[00:01:27.293,579] <inf> app: Flushing UDP
[00:01:27.742,919] <wrn> modem_monitor:    State: Active
[00:01:33.686,065] <wrn> modem_monitor:    State: Idle
...
[00:04:10.787,231] <inf> app: Flushing UDP
[00:04:11.278,381] <wrn> modem_monitor:    State: Active
[00:04:16.875,518] <wrn> modem_monitor:    State: Idle

Is this expected behaviour? I have no idea.
I would have thought that `NO_DATA` would result in a quicker transition to Inactive since it specifies tighter constraints (No more TX or RX, instead of only no more TX), but that doesn't appear to be true.
But if it is expected that both options have the same behavior, why would they be split into separate options at all?

[Achim Kraus]

Maybe Please consolidate nRF9160 documentation about CP-RAI and AS-RAI

Or NAS-RAI (Non-Access Stratum RAI) used on NB-IoT with nRF9161/51

So, it will important, if you use a nRF9160 (mfw 1.3.7) or a nRF9151 (mfw 2.0.4). The later reports, which RAI variants are supported. for the first one, it may get only visible, if you check the time on your own. In my experience, it takes about 2s, so 5s maybe rather just the idle time of the network provider. Did you also check without RAI to see, what you get then?

>  `NO_DATA` would result in a quicker transition

Not really verified, but AFAIK NO_DATA also requires a radio message to be sent.

[JordanYates]

I'm not looking for how to check network support, that is not hard information to get (there is a builtin event for it).
Using `LAST` or `NO_DATA` definitely appears to drop off faster (5s rather than 12s), but that also doesn't answer the question.

> Not really verified, but AFAIK NO_DATA also requires a radio message to be sent.

And that's sort of the point, there is no documentation that I can find about what the expected behavior is. The source has the conditions under which the various options should be set, but nothing about what the actual result is.

[Achim Kraus]

Let me start with, I'm just an other user and not an employee if Nordic ;-).

> I'm not looking for how to check network support, that is not hard information to get (there is a builtin event for it).

Then you use a nRF9151 with mfw 2.0.4. If a older one, the nRF9160, is used, there is no such event. 

> And that's sort of the point, there is no documentation that I can find about what the expected behavior is.

In the past, such things where frequently linked to some GSMA "low level spec.", but we will see, if you get some better answers from Nordic. 

[Simon D-M]

Hi,

We have a bit of documentation about RAI (link).

Basically the "NO_DATA" is not tighter that "LAST":

• "NO_DATA" : no more TX and RX planned
• "LAST": no more TX after the next transmission. (RX still available)

So in some cases "NO_DATA" could be faster to release however the decision to release the connection is made by the network provider, not the UE.

I think that the 5sec delay in both is due to the network provider waiting 5 seconds, which makes both option behaving in the same way.

If you want to really confirm it, you can send us a raw modem trace (.bin or .mtrace), and we should be able to see this behavior in the trace.

Best regards,

Simon