NV2 QoS: question regarding control frame timeout

We are testing NV2 QoS feature on RouterOS 6.1 with:
• RB433AH with 2xR52Hn and external antennas as AP
• SXT 5HnD as wireless clients (12 per R52Hn)

NV2 QoS parameters on AP are configured in the following way:
• Queue count: 4
• QoS: frame priority

AP wireless interfaces are working in „ap bridge” mode while wireless interface on clients is working in „station bridge“ mode.

On both AP and client side only the VoIP traffic is matched by the simple mangle and bridge filter rules which set the traffic priority to 7 (for nv2 QoS priority the highest is better – confirmed by lab tests!)

In AP logs I can often see control frame timeout disconnects:

00:0C:42:89:23:35@w1: disconnected, control frame timeout

Is it possible that nv2 control frames are not prioritized by default and therefore under some circumstances VoIP traffic “starves” the queue thus disabling nv2 control frames to pass which causes link outages?

If that is the case, is there a way to match nv2 control traffic? Example?

With NV2 protocol it is prone to co-location interference or noise and solution we found very effective to stop control frame timeouts is to apply RF screening to AP’s or purchase AP’s with good RF screening?

Thank you for suggestion on “control frame timeouts” solution. I couldn’t agree better that NV2 protocol does not like the noise so much.

Do you maybe have any idea regarding my original question, the relation between NV2 QoS rules and “control frame timeouts”? Is there a possibility that high priority VoIP traffic starves available queue and doesn’t left any space in it for nv2 control protocol frames? Is NV2 control traffic prioritized by default? If not, is there a way to do it? Question asked in a bit different way: will link be broken if you try overload its capacity?

I can only speak from experience with control frame timeouts caused by co-location interference, where the AP’s could not communicate with clients within the specified time slot?

As regards your question I max out links for throughput tests and don’t get control frame timeouts, the usual drop in CCQ occurs but I would check RB433AH CPU is not set to low frequency, had a lot set to 300 when they could be set to 680Mhz – higher was available but not recommended I believe.

Also as general practice I uninstall packages and disable services not required by the router

When using two radio cards onto one routerboard self interference will be a major issue, and especially as throughput increases,

I use a simply rule which works good for me and that is 1 radiocard = 1 routerboard

+1

Shouldn’t be RB433AH designed for 3 cards? Can this really be an issue?

Yes - strange but true you cannot use a board designed for 3 cards with 3 cards it causes self interference unless perhaps using 900Mhz, 2.4 and 5.8GHz radio cards but certainly not the same frequency band?