Link Layer Adaptation

Hello i am using a modem router from my ISP because it has VOIP...and it connects to ISP with ppp0e... i have enabled passthrough in this modem router and i have connected a TP_LINK WR1043nd-v3 with openwrt 18.06.0 and with WR1043nd it makes a ppp0e connection to.
So in Link Layer Adaptation-->Per Packet Overhead what value it must have?

Hard to tell, is this an adsl, vdsl2, or docsis/cable link?

It is vdsl2...30Mbps download and 3Mbps upload...so what value in packet overhead it must have?

So for vdsl2 you will have at least:
VDSL2 (IEEE 802.3-2012 61.3 relevant for VDSL2): 1 Byte Start of Frame (S), 1 Byte End of Frame (Ck), 2 Byte TC-CRC (PTM-FCS) = 4 Byte

COMMON (VDSL2/Ethernet): 6 (dest MAC) + 6 (src MAC) + 2 (ethertype) + 4 Byte Frame Check Sequence (FCS) = 18 Byte

Total: 22 Byte

plus potentially:
PPPoE: 2 Byte PPP + 6 Byte PPPoE = 8 Byte

VLAN (IEEE 802.1Q): 4 Byte VLAN = 4 Byte

Total 34 Bytes if using pppoe and VLAN

(There might be other optional overheads I do not know about). Typically an ISP will have some opaque pieces of documentation at hard to reach places that might help elucidate these questions if one only would find them...

Best Regards

1 Like

What's the overhead on a docsis 3.0 link. Vergin UK I herd it was 18. Is there a way to test and see what the overhead is? Then people can make shure of using the rite settings.

After Googleing I found this:
https://lists.bufferbloat.net/pipermail/codel/2016-June/002325.html

Hi tapper,

we later cleared up the confusion with Greg. The initial question and answer are not wrong, but irrelevant for end users, as the end user bandwidth shaper assumes an 18 bytes overhead per packet. The overhead discussed with Greg is closer to the real overhead of a docsis system, but that simple never becomes an issue in a docsis network. Hope this helps.

So what are you saying? I should stick with 18 or leave the field empty?

Yes, sticking to 18 seems to be the best advice.

@moeller0 Do virtual interfaces add any additional overhead? e.g. eth0.2 for wan, eth0.1 for lan (in my case)

A vlan tag will add 4 byte to the overhead (see https://en.wikipedia.org/wiki/IEEE_802.1Q), but if the vlan tag is only used between your router and the modem (so not between the modem and the ISPs upstream device) it is not relevant for your wan shaper. In general the idea is to describe and account for how the packets look on the narrowest part of the network path as that will be the speed limiter, transparent overhead before that should not count....

That clears things up a bunch.

@moeller0
I'm using VDSL and PPPoE connection requires a vlan tag.
What if my bridged modem already sets the vlan tag? Would the proper overhead be 30 instead of 34?

No, you need to explicitly tell the shaper the total overhead at the bottleneck-link, even if the shaper is instantiated on the vdsl modem itself you still need to explicitly tell it which overhead to account for.
That said qdiscs other than cake will automatically add some of the known overhead, which in most cases is worse than not adding any. Cake will, unless asked to act differently, assume that overhead is intended to be calculated to be on top of the ethernet payload portion...
Hope that helps...

1 Like

I'm using Cake > Layer_Cake and my interface is pppoe-wan, so 34 should be fine for my setup.
Thanks for the clarification.