Have our first 820C link up and trying to commission it for a customer. We are running 512QAM at 550 x 550 for each radio chain. Running 9.7 on our radios. Have LAG configured 1 group for the ethernets and one for the radio interfaces. Staying in good modulation no issues there. Ethernet interfaces are both ethernet, one off the POE port and one off the ETH2 via copper SFP. LAG group built in a Juniper switch on each sides.

We are trying to commission the link. Trying to get around 850M total through it. What we are seeing is that we can get about 540M through and then we get about another 200 through running 2 separate sets of test sets. Sometimes we see 550 on one chain and then we only see 500M in one direction over the test sets. Sometimes the direction we are getting 500M reverses. We tried dropping one of the ethernet interfaces and testing and we are seeing the same thing.

We’ve reviewed the release notes for 9.7, which Cambium support recommended we be on and can’t find any issues describing what we are seeing. We are using 4 x Mikrotik CCR’s for our testing, also tried Lanforge and we were seeing the same thing.  We are using OOB management and a single PIPE interface for a straight L2 connection on the data ports. The PIPE interface contains just the LAG groups for the Ethernet and Radio.

Any ideas? We’ve had a few different folks from Cambium support look over the config and they don’t appear to see any issues each time they look.  Cambium also configured the LAG with us as it's the first link we've done in this configuration.  We do have a ticket open, just looking to see if there's anything we can try as we have been unsuccessful in getting more throughput through the radios than the 540 - 550M we are seeing on each RF link.


How do you do the LAG throughput testing? Please notice that LAG relies on some address hashing machanism so you cannot garantee that the traffic can be load balanced among the two links. If you are only using one traffic stream, the traffic load will be directed by the switch to only one of the ports so that the other port gets nothing. For a LAG to achieve good load balance, you need a big number of data streams with nice distribution of the traffic addresses.

Hope this help.



1 Like


Thanks for the reply!  We were doing testing with 2 streams.  The odd thing we've found is if we have the ethernet LAG and both ports turned up as well on each side we get very poor results like I mentioned.  When we only have one of the ethernet interfaces up, we can get consistently 950 - 960M through the link, but as soon as we turn LAG back on we get poor results.  We have the Radio interface lagged as well.  

Assuming that you are using dual carriers. What you need to do is to setup two PIPEs. You dont need to do any LAG at the PTP820C configuration at all. What you need to do is to have the external switches take care of the LAG.