Forum

Hi folks,

Curious to know if anyone is deploying LWAPP Auto-RF just for Dynamic Channel Allocation (DCA), not for Dynamic Transmit Power Control (DTPC). We have a dense deployment of 1231 APs (ten floors of a hospital with a 4-leaf clover design, with "pods" at each leaf having 2 APs each) and have enabled DCA only, with a static Tx Power constant of TPL3 (31mW). Note: the original IOS AP setup was at 20mW and we intend to lower the TX Power to TPL4 (FYI - 1230s and 1240s do not have equivalent dBs for the same power levels in LWAPP). Its also our understanding from the Cisco docs that AP neighbor beacons sent at a hard-set 100mW are controlling DCA behavior regardless of the user-configured transmit power.

We are finding that DCA is going nuts in this environment (12-24 channel changes/day, on several floors of APs at once). It seems to never stabilize. And it impacts both APs in a coverage area: each floor "pod" loses both APs (due to client de-authentications prior to the channel change) at the same time. Is it acceptable to enable Dynamic Channel Allocation but not enable Dynamic Transmit Power Control in this way? Could this be contributing to the DCA activity?

Also, if clients are not de-authenticated when LWAPP APs go off-channel to scan for Interference/Rogue APs/Noise, that clients have to be de-authenticated when the Dynamic Channel Allocation process chnages the channel? Aren't both processes doing the same thing (changing channels), and taking the same time to accomplish it (50-60ms)? It seems like de-authenticating the clients is like throwing the baby (client) out with the bathwater (channel). If the client is sustainable during the off-channel scans how less sustainable are they when the channel plan changes?

Sorry my 2nd question is why does it de-authenticate clients when it changes channel when it doesn't have to when it goes off-channel to scan? Is there soemthing in the 802.11 standard that dictates this?

Deauthentication is a normal process of 802.11 roaming. If you look at the capture you will see the reassociation (it may be just an association frame) occur then the deauth. The client device doesn't need to deauth to change channels for probing because it isn't actually wanting to disconnect from it's AP, its just looking for a better AP.

Hi Bruce Johnson,

Just curious how much noise, interference and user mobility is in that environment?

Are you using wireless voice applications and numerous VLANs?

Are there Rogue devices moving in and around often?

You might have read this already "DCA is looking at this:

http://www.cisco.com/en/US/products/ps6306/products_white_paper0900aecd802c949b.shtml

The Cisco Wireless LAN Controller examines a variety of real-time RF characteristics to efficiently handle channel assignments. These include:
?¡é?€?¡é Access point received energy-This is the static topology of the network; this feature optimizes channels for the highest network capacity.

?¡é?€?¡é Noise-This limits signal quality at the client and access point. An increase in noise reduces the effective cell size. By optimizing channels to avoid noise sources, the Cisco Wireless LAN Controller can optimize coverage while maintaining system capacity. If a channel is unusable due to excessive noise, that channel can be avoided.

?¡é?€?¡é 802.11 interference-If other wireless networks are present, the Cisco Wireless LAN Controller will shift the usage of channels to complement the other networks. For example, if one network is on Channel 6, an adjacent WLAN will be assigned Channel 1 or 11. This increases the capacity of the network by limiting the sharing of frequencies. If a channel is used so much that no capacity. is available, the Cisco Wireless LAN Controller may choose to avoid this channel.

?¡é?€?¡é Utilization-When this is turned on, the capacity calculations can consider that some access points are deployed in ways that carry more traffic than other access points (a lobby versus an engineering area, for example). More emphasis is given to assigning channels to the access points that require the most bandwidth.

?¡é?€?¡é Client load-Client load is taken into account when changing the channel structure to minimize the impact on the clients currently on the WLAN system. The Cisco Wireless LAN Controller periodically monitors the channel assignment in search of the "best" assignments. Change only occurs if it significantly improves the performance of the network or corrects the performance of a poorly performing access point.

The Cisco Wireless LAN Controller combines the RF characteristic information with intelligent algorithms to make systemwide decisions. Conflicting demands are resolved using soft decision metrics that guarantee the best choice for minimizing network interference. The end result is the optimal channel configuration in a three-dimensional space, where access points on the floor above and below play a major factor in an overall WLAN configuration.

Thanks guys,

We have read a lot of Cisco docs and talked to their folks but unfortunately Cisco has not come forward with many quantitative values to these features like prioritizing improved channel selection for APs under load and how to modify them to suit our environment. AFA interference and noise, te WCS is not reporting any issues (and we've seen no interference using Cognios SA. but we may need more long-term data which will be available in 4.1 code, which allows a Cognio laptop to send traps to the WCS), which leads me to believe its responding to its own 100mW Neighbor messages with so many high RSSI neighbors that it can't get out of its own way.

The LWAPP infrastructure de-auths clients prior to a chennel change, but to my understanding it does not do so when assessing the RF for noise and interference as the basis for this channel change (in 50ms slices within a default monitoring window of 300 seconds). My question is how is the RF assessment phase (going off-channel to scan) any different from the channel change it makes when it thinks it can improve the signal quality of a -65 RSSI or higher AP neighbor AP by 5dB (this is Cisco's stated goal of the algorithm for making a channel change).

Do your site survey with the same antennas you'll be using for the deployment and your RF characteristics will be the same when the APs are LWAPPed and joined to the controllers.

Except that the controller's interpretation may not be as the AP/antenna layout was designed.

Go to this link read this post . It might help understand this anomaly. Scroll down to topic on Cisco Ripple Effect with RRM. This blogger ran into issues that may be similiar.

http://www.hubbert.org/atom.xml

Thanks Compughter,

This is interesteing, and certainly what we're seeing. The author is suggesting its due to rogues (and we have thousands of these across our footprint). However I have seen any of these hit the neg 85 threshold necessary to trigger DCA. Perhaps the interference floor is lower than neg 85, but neither the controllers nor the WCS are alerting us to interference issues, which I'd expect if they were casuing this.

The Avoid AP load metric is somewhat counter-intuitive, in that it says it will change channels to provide better signal for APs under load; I thought this meant it would avoid changing channels on APs with clients, but it seems it may actually *prefer* changing these APs if it thinks it can improve their signal. I'll be unchecking this to see if it changes the DCA frequency.

Thanks for the link,

Bruce Johnson