RFS7000 Multi-link trunk - load balance

[Phil_storey]

Hi

    We have two RFS7K - currently it is set to tunnel , so all wifi traffic comes back to the controller, is it possible to channel bond or create an MLT into the network stack so the throughput is increased

There is only one port connected on the RFS which is GE1

[Phil_storey]

Hi

  Thankyou both very much for your input, @Christopher Frazee " We strongly advise configuring WLANs bridging mode local "  I Did try flipping the WLANS over to locally bridged, setting the AP GE port to be a trunk , but wasn't sure if you still need to keep the virtual adapters for the wlan to vlan MAP.

Would it still work oK till such time as we do flip from tunnelled to locally bridged ?

Phil

[vanelm · Accepted solution]

Hello Phil,

  Load balancing strategy depends on traffic profile. In case of single default gateway in particular VLAN - mac method should be relevant. The idea here is to keep a single flow path unchanged.
  Regarding balancing RFS units - please keep in mind that tunneled VLANs is anyway balanced in the cluster by default - i.e. half of VLANs are bridged on every cluster node (in your case). This apparently means that user traffic also tunneled between nodes to reach the target bridge  or DIS depending on there AP adoption tunnel landed (please see “sh mint dis detail”). This basically means that communication between nodes is also important and should be effectively organised. Like using  mentioned stack of switches.
  Stack is totally good as long as it appear as a single unit and support LACP

port-channel load-balance src-dst-?
  src-dst-ip   Source and Destination IP address based load balancing
  src-dst-mac  Source and Destination Mac address based load balancing
 

Regards,

  Misha

[Christopher_Fra]

I use source-destination IP address for port aggregation/EtherChannels with Wing switch and Cisco L2, with no issues. I would definitely have the 2 RFS controllers physically connected into different L2 switches for redundancy purposes. The RFS7 has 4 ge ports that are on their own back plane, allowing a total of 4GB of aggregated throughput.

We strongly advise configuring WLANs bridging mode local and carry the VLANs on the L2 switches that the APs are physically plugged into and the on the APs. The RFS would only have the management/adoption VLANs. This would mitigate any bottleneck concerns at the RFS.

[Phil_storey]

Hi Misha

Many thanks for the reply 

   are there any caveats to this ? i.e if you have a primary and backup unit ?  is it better in your opinion to use IP or MAC for the destination of the network switch ?

The rfs is into a Nortel stack of 4 switches, So I’m think of using a ports from different switches in the same stack, that way if we loose a switch it will keep running

regards

Phil

[vanelm]

Hi, 

In that case I would recommend consider a local bridging first - perhaps switches support VLANs. After that done, consider RFS7k replacement with VX9000 and WiNG upgrade to the latest-what-is supported-by-aps, like 5.9.5 - this brings security fixes (like KRACK) and some useful features.

  Finally yes, RFS7k support port channels and LACP

#self

interface port-channel 1

 port-channel load-balance src-dst-mac

 switchport mode trunk

 switchport trunk allowed vlan 1,20-120

interface ge 1

  channel-group 1

interface ge 2

  channel-group 1

Regards, 

  Misha