Interface Groups - interfaces unevenly distributed

[Anthony Poli]

                   We're leveraging interface groups to break up broadcast domains into 24 bit nets while still providing ample IP space to clients associated to a specific WLAN.  As I understand it the WLC should round-robin the interfaces to clients as they associate, however we are seeing this unevenly skewed where some interfaces are being overutilized and others underutilized.  I've attached a screenshot of our DHCP server which demonstrates this behavior for the networks provisioned to this WLAN.

Has anyone else experienced this and is there any way to correct it? 

Thanks!

Anthony

[Scott Fella · Accepted solution]

Well the WLC will calculate which interface it will use depending on the MAC address of the client. If you have added interfaces to the group, that can cause a skew also. In order to start fresh, you can remove and recreate the interface group or reboot the WLC. I wouldn't work about it since the WLC will mark the interface dirty if a client fails to revive an IP address.

Sent from Cisco Technical Support iPhone App

[Rasika Nayanajith]

Hi Anthony,

This is normal behaviour from WLC 7.2.x onwards when you using "interface groups". Refer this link for more detail & listed the key points of algorithm used for IP allocation.

http://www.cisco.com/en/US/products/ps10315/products_tech_note09186a0080bb4900.shtml?referring_site=smartnavRD

1. When a client associates to a WLAN on a controller, an index is calculated based on the MAC address of the client and the number of interfaces in the interface group using a hashing algorithm.

2. Based on this index, an interface is assigned to the client.

3. Whenever this client joins the controller, the hashing algorithm always returns the same index and the client is assigned to the same interface.

4. If the interface is “dirty”, then a random index is generated and the interface is assigned based on that random index.

5. If that interface is still dirty, then a fall back to round robin implementation occurs.

If you are using 7.x code below 7.2 then it is more or less round-robin algorithm & you should see equally utilized subnets. But it has some drawbacks & cisco modified it in 7.2 onwards.

http://www.cisco.com/en/US/products/ps10315/products_tech_note09186a0080b78900.shtml

HTH

Rasika