Link Aggregation (IEEE 802.1AX, formerly 802.3ad — commonly called LACP, LAG, “bonding,” or “port-channel”) lets two network devices combine several physical links into one logical one for extra throughput and automatic failover. ePMP AX now supports carrying LACP traffic transparently across a wireless hop, so a wireless link can serve as one member of a LAG between two third-party devices.
No configuration needed on the radios. The ePMP AP/SM pair does not participate in the LACP negotiation and has no LACP settings to enable. It simply bridges the IEEE 802.3 Slow Protocol frames (EtherType 0x8809, destination MAC 01:80:C2:00:00:02) end-to-end, exactly like it bridges any other Ethernet frame. From the perspective of the two devices running the LAG, the wireless bridge behaves like a plain cable.
Validated topology. Two LACP-capable devices (“Router A” / “Router B”) were bonded over two independent member links. See the diagram below.
Known limitation — managing the radios through the bond. Because LACP’s hashing algorithm picks a member link purely by traffic hash, without knowledge of which physical path actually leads to a given end device, management traffic aimed at the ePMP AP’s or SM’s own IP address can be hashed onto the member that doesn’t lead to it. This is a property of link aggregation itself, not something the ePMP or the bonding device gets wrong.
Recommendation: manage the ePMP AP/SM out-of-band, or address it over the specific physical member port it is actually attached to — don’t rely on reaching its management IP through the bonded/LAG interface.
I thought the “Channel Bonding” feature developed for the PTP550E—which works very well—could be carried over.
We still have some PtP backhauls using the PTP550E, and they never fail; the only thing that affects them (at times) is interference, which reduces capacity, but we use DCS and ACS to handle that.
I had always hoped that feature would be migrated to the 3000 series and now the 4000 series, but from what I can see, that won’t be possible.
Regarding LACP: when you say management is lost, how can we monitor the devices—checking their status, capacity, and interference levels? That part isn’t clear to me.
The PTP 550 has three radios under the hood: two for Tx/Rx and one dedicated to the spectrum analyzer. The bonding between the two Tx/Rx radios is our own feature, so it’s entirely under our control. The E3K/E4K has only two radios, one for Tx/Rx and one for the spectrum analyzer, so there’s nothing to bond.
As for LACP, customers kept asking for it, so we decided to go ahead and add it. That said, we don’t participate in the negotiation at all; we simply forward the control frames. Proper radio management and monitoring isn’t possible within the LACP design, since the protocol doesn’t expect anything sitting between the link partners. It was designed for wired networks. For that reason there’s no configuration we can offer here.
That said, I’d encourage you to experiment. If you find a method that works reliably, please share it with the community.
Estoy intentando configurar LACP con dos antenas en cada extremo pero al momento de de configurar el portchannel en modo active se van a modo suspended las interfaces, tal cual está el diagrama la única diferencia es que el modelo es cambium ptp 550 E, este upgrade de la versión 5.12.1 no estara disponible para este modelo tambien?
