Dynamic interfaces, one VLAN, multiple physical interfaces?

Answered Question

Hi

We are just getting started with a 5508 WLC and WCS. We can already see this is a big improvement over our present setup with autonomus AP's. We are also implementing some 11n AP's as part of this.

I think I understand the multiple AP-manager interfaces and the way AP's load balance over these. But I dont understand the way the client traffic should be load balanced.


The goal is to have one WLAN. When I create this I select a dynamic interface (and thus a VLAN for the clients). But this VLAN is bound to a physical port (with a backup port). So from my understandig the client traffic wil go from the AP's to the controller on multiple interfaces, but from then on will go out to the servers on just one interface?

Thats not what I want - I would like the traffic out on the VLAN to the servers to be distributed on more than one port. How do I do that? Do I then have to use LAG?

Regards,


Kaj

I have this problem too.
0 votes
Correct Answer by Matthew Fowler about 6 years 12 months ago

If the buildings are close enough to cause a roaming event, the client will maintain it's previous address and remain on the previous VLAN i.e. a layer 3 roam will take place. If it takes more than the User Idle Timeout (default 300 seconds) to move between buildings, there will be no client table entry in the WLC so the client will get a new address and be put in the new VLAN.

So, a roam should not take any longer, but you will see clients in building B with addresses from building A and vice versa if a client takes less than 300 seconds to move between buildings.

And yes, LAG between multiple switches in the same stack is supported.

Correct Answer by dancampb about 6 years 12 months ago

Correct, if you only have one WLAN you probably would only have one dynamic interface, unless you are using AP Group VLANs.

  • 1
  • 2
  • 3
  • 4
  • 5
Overall Rating: 5 (2 ratings)
Loading.
dancampb Mon, 12/07/2009 - 09:45

OK, let's see if I can explain this.  If you aren't running LAG, for every physical port you have cabled up you need to have an ap-mgr interface created and assigned to that port.  The AP's will automatically load-balance over all the ports.  On the controller you can see which port the AP is talking through.  So all traffic between that particular AP and the controller will go over that one port, unless there is a failure then it would move to another port.

Same is basically true with the client traffic.  You assign the dynamic interface to a primary port and a backup port.  All the unencapsulated traffic to/from the clients on this WLAN would go in and out of the port mapped to the dynamic interface.  I say unencapsulated since the traffic to/from the AP for that client would be through the ap-mgr interface.

The idea of load-balancing in a non-LAG setup is dyanmic interface 'A' is assigned to port 1, interface 'B", to port 2, etc.   This way all the traffic isn't flowing through one port which is being heavily utilized when the other ports aren't untilized much at all.

LAG is a way takes care of all the load-balancing automatically.  LAG is simply a port-channel on the controller.   So you enable LAG on the controller and setup the etherchannel on the switch and all of the traffic to and from the controller is load balanced over all of the physical connections.  Also with LAG enabled you only need to have one ap-mgr interface.

Thanks for the help dancampb. A few more questions 

OK, let's see if I can explain this.  If you aren't running LAG, for every physical port you have cabled up you need to have an ap-mgr interface created and assigned to that port.  The AP's will automatically load-balance over all the ports.  On the controller you can see which port the AP is talking through.  So all traffic between that particular AP and the controller will go over that one port, unless there is a failure then it would move to another port.

Ok, so far I seem to understand this right.

You assign the dynamic interface to a primary port and a backup port.  All the unencapsulated traffic to/from the clients on this WLAN would go in and out of the port mapped to the dynamic interface.

When I create a WLAN I choose a specific dynamic interface for the WLAN. So all traffic for this WLAN will go in and out the specific port for the dynamic interface. So a WLAN that spans several 11n APs will be able to hit a limit on the 1Gbit on the physical port?

I can create multiple dynamic interfaces, but that would only help if I also have multiple WLANs?

Kaj

Correct Answer
dancampb Mon, 12/07/2009 - 10:29

Correct, if you only have one WLAN you probably would only have one dynamic interface, unless you are using AP Group VLANs.

Thanks a lot for your help


You are probably hitting something important. With around 400 concurrent users I should probably use AP group VLANs. I will have to read up on that. I can easily divide into two AP groups (two buildings). How seamless is client roaming between the VLANs when the client will have to change IP address?

Or I should use LAG. LAG connected to more than one swith is no-go, but connected to a stack of 3750G would be OK?


Regards,


Kaj

Correct Answer
Matthew Fowler Mon, 12/07/2009 - 17:02

If the buildings are close enough to cause a roaming event, the client will maintain it's previous address and remain on the previous VLAN i.e. a layer 3 roam will take place. If it takes more than the User Idle Timeout (default 300 seconds) to move between buildings, there will be no client table entry in the WLC so the client will get a new address and be put in the new VLAN.

So, a roam should not take any longer, but you will see clients in building B with addresses from building A and vice versa if a client takes less than 300 seconds to move between buildings.

And yes, LAG between multiple switches in the same stack is supported.

Actions

This Discussion

Related Content