MaxRetxThreshold in LTE RLC-AM

[Luis Diez]

Hello all,

I'm working with the RLC-AM of LTE and some questions have popped up.

The RLC-AM uses a counter for the maximum number of retransmissions for a given RLC PDU, but in fact that counter is not used. In the "DoReceivePdu" method the counter is checked just to show a message, but no action is carried out (this applies to different releases including 3.21). I think that the standard only says that it must be communicated to the upper layers (rather cryptic). In the end the result is that for a some seconds of application with bad channel conditions (so with small size of the transmission opportunity) the RLC protocol continues sending traffic, due to the retransmissions, much longer after than the application gets closed. Indeed, the sender buffers remain with data tens of seconds after the last PDCP PDU is sent down to the RLC sender.
Now the questions:
Is there any way to avoid this behaviour (other than shutting down the simulation :) )?
Would it be possible to discard PDUs and let the upper layers to solve it, if necessary?

Thanks in advance.

BR,
Luis

[Nicola Baldo]

If I recall correctly, the standard says you should trigger Radio Link Failure if the retx limit is reached. Unfortunately, RLF is not implemented yet.

[Luis Diez]

Hello again,

Thank you for the response, Nicola.
You must be right in that RLF must be initiated, unfortunately TS 36.322 says nothing about RLF. I mention the technical spec. because the current implementation seems to follow it to the letter. It must be one of those things let to the implementers.

On the other hand, I've seen that the RLC-AM implementation foresee the possibility of losses (at some point in the RLC SDU re-assembly). Do you (any) see any major issue if once the MaxRetxThreshold is reached the PDU is discarded? I mean, the transmission window is moved forward as if an ACK would have been received.

[Nicola Baldo]

TS 36.322 says that the RLC should notify upper layers when the Retx threshold is reached.
Then TS 36.331 says that the RRC should consider this notification as an indication of RLF.

I don't see any part of TS36.322 saying that the RLC should or could autonomously decide to drop a PDU when the retx threshold is reached.
That said, you are free to do whatever you want with your local copy of the code :-)

[Luis Diez]

Hello Nicola,

thank you for that information, this really sheds light in my understanding on how things should go on and why that behaviour happens.
I just wanted to avoid the so long simulations I mentioned in my first comment. Anyway, I'm afraid I'll have to work it around somehow.
Sorry for turning the discussion to standards :)
Thank you again for your time

BR,
Luis