LWAPP vs Autonomous survey

[jsmbrown]

Does it take more LWAPP APs than Autonomous APs for a controller environment to function correctly with Auto RF? I am specifically concerned with channel assignment working properly, more than the self healing ability of powering up to cover APs that might go down.

I would appreciate any document links supporting either opinion.

Thanks,

Joe

[tkhan]

I would say, "no", given that the controller based technology provides more intelligence for coverage than autonomous.

[SJessulat_2]

My experience is, if you do a site survey with autonomous APs, you can easily use these AP locations for LWAPP-APs.

The controller uses the same maximum power settings as the autonomous APs.

For channel assignment: The Controller based Auto RF function assigns the Channels dynamically, so usually you dont have to worry about channel assignment. But in case of any problems, you can still override the controller settings manually.

[Rob Huffman]

Hi Joe,

Just to add a note to the good info from Tyron and Sebastian. When we converted from an Autonomous deployment to LWAPP we did not change the AP locations or add AP's to cover existing areas. The DCA only kicks in when an AP can hear it's neighbors within an RF-Group. So if the AP's are not close enough to hear each other then DCA is moot anyways. If they are close enough then the algorithim is used to select non-interfering channels and you are good to go :)

Dynamic Channel Assignment Algorithm

The DCA algorithm, run by the RF Group Leader, is applied on a per-RF-Group basis to determine optimal AP channel settings for all the RF Group's APs (each set of APs who can hear each other's Neighbor Messages, referred to in this document as a logical RF sub-Group, has its channel configuration done independently of other logical RF sub-Groups due to the fact that signals do not overlap). With the DCA process, the leader considers a handful of AP-specific metrics that are taken into account when determining necessary channel changes. These metrics are:

Load Measurement-Every AP measures the percentage of total time occupied by transmitting or receiving 802.11 frames.

Noise-APs calculate noise values on every serviced channel.

Interference-APs report on the percentage of the medium taken up by interfering 802.11 transmissions (this can be from overlapping signals from foreign APs, as well as non-neighbors).

Signal Strength-Every AP listens for Neighbor Messages on all serviced channels and records the RSSI values at which these messages are heard. This AP signal strength information is the most important metric considered in the DCA calculation of channel energy.

These values are then used by the Group Leader to determine if another channel schema will result in at least a bettering of the worst performing AP by 5dB (SNR) or more. Weighting is given to APs on their operating channels such that channel adjustments are made locally, dampening changes to prevent the domino effect whereby a single change would trigger system-wide channel alterations. Preference is also given to APs based on utilization (derived from each AP's load measurement report) so that a less-used AP will have a higher likelihood of having its channel changed (as compared to a heavily utilized neighbor) in the event a change is needed.

Note: Whenever an AP channel is changed, clients will be briefly disconnected. Clients can either reconnect to the same AP (on its new channel), or roam to a nearby AP, which depends on client roaming behavior. Fast, secure roaming (offered by both CCKM and PKC) will help reduce this brief disruption, given there are compatible clients.

Note: When APs boot up for the first time (new out of the box), they transmit on the first non-overlapping channel in the band(s) they support (channel 1 for 11b/g and channel 36 for 11a). When APs power cycle, they use their previous channel settings (stored in the AP's memory). DCA adjustments will subsequently occur as needed.

From this excellent RRM doc;

http://www.cisco.com/en/US/tech/tk722/tk809/technologies_tech_note09186a008072c759.shtml#dcaa

Hope this helps! And makes some sense.

Rob

[jsmbrown]

Thank you Rob. I have read the RRM document many times. I am in the process of troubleshooting a couple of buildings where wireless was implemented. There is plenty of signal throughout the building, but several APs in the same area are choosing the same channel and causing significant co-channel interference. Manual override of channels is not an option for this customer. They were sold on the idea of auto-rf and rightly so expect it to work. If we change DCA sensitivity to high, it seems to move to better channel allocation. The problem then though is that the APs are performing more channel changes than the customer wants. So I am trying to learn all I can about DCA. One exert from part of the document you posted that I would like an explanation on is - These values are then used by the Group Leader to determine if another channel schema will result in at least a bettering of the worst performing AP by 5dB (SNR) or more -

Can anyone explain to me what this means? Please include an example and doc links if possible

Thanks,

Joe

[jsmbrown]

Thank you Rob. I have read the RRM document many times. I am in the process of troubleshooting a couple of buildings where wireless was implemented. There is plenty of signal throughout the building, but several APs in the same area are choosing the same channel and causing significant co-channel interference. Manual override of channels is not an option for this customer. They were sold on the idea of auto-rf and rightly so expect it to work. If we change DCA sensitivity to high, it seems to move to better channel allocation. The problem then though is that the APs are performing more channel changes than the customer wants. So I am trying to learn all I can about DCA. One exert from part of the document you posted that I would like an explanation on is - These values are then used by the Group Leader to determine if another channel schema will result in at least a bettering of the worst performing AP by 5dB (SNR) or more -

Can anyone explain to me what this means? Please include an example and doc links if possible

Thanks,

Joe

[Rob Huffman]

Hi Joe,

I see your dilemma here. Have you talked to the custmer about using the DCA initially to set the channels and then turning it off so that there are not changes happening so often. It seems to me that a good "stable" service is most important at this time.

Their concerns with RRM in general could be pushed at a "higher" level with your Cisco SE and Sales team.

Just a thought,

Rob

[jsmbrown]

I could suggest that, but I have a feeling they will want all features and functionality to work all the time.

The local Cisco SE and sales team are saying the customer may not have enough APs. Before I accept that, I was wanting to know if more APs are needed for LWAPP than Autonomous. Everything wirelessly works great if channels are hard set. So to say more APs are needed just for RRM to function seems like a poor design. That is why I was asking if others found it necessary to add more APs to LWAPP vs autonomous and I am really curious if that is documented anywhere on Cisco's site. I have yet to find that documented, in fact everything I see so far points to the idea that you do not need anymore for LWAPP.

[tkhan]

Hey Joe, we went through a similar migration to Rob's. We had about 1300 AP1230Gs running IOS that we converted to LWAPP, in about 78 locations across the country. We didn't add a single AP during these migrations. We did have some issues with early versions of AUTO RF where it ran every 10 minutes and turned out to be disruptive to our clients (power and channel changes). With later versions of code, which we run 4.2.130.0 today, these issues went away when using the default "medium" sensitivity setting for AUTO RF. In a highly changing RF environment, you can either do as Rob suggested, turn AUTO RF off and use static values or change AUTO RF sensitivity to "LOW". You can also run an "ON DEMAND" event when your environment is saturated with RF to allow AUTO RF to make the best assumptions for power and channel setting for your APs.

[jsmbrown]

Thanks for the information. How dense were the APs to each other?

The customer in this case told us many times they do not want an AP in every room and designated specific areas where more APs should be placed for density of wireless clients. They specified basic wireless coverage, but not to be designed for density. Therefore our APs were placed in about every other room and most are at full power. The problems we are having are only in specific areas of the buildings and not all throughout. One example is 9 out of 11 APs on channel 11 in one hallway. In this case, I looked at the AP neighbor list of the APs to determine what neighbors they see and at what power levels. 2 of the worst offenders were on the same channel as their strongest neighbors. The neighbor list was as follows:

I will call the 2 trouble APs - AP 1 and AP 2.

AP1 channel 11

AP1 neighbors:

AP2 -69 11

AP3 -76 11

13 total neighbors all other neighbors were above -80 dBm ( my understanding is only neighbors below -80 are considered by RF Group Leader for DCA )

AP 2 channel 11

AP 2 neighbors:

AP1 -72 11

AP4 -58 1

AP3 -77 11

AP7 -76 11

12 neighbors total - others are greater than -80 dBm

If we change DCA sensitivity to high, it starts using channel 6 more and APs next to each other are using different channels. It is almost the same assignment as I would use if I manually set them.

[SJessulat_2]

The larger the deployment the longer it takes for Auto-RF to get into a stable state, so that channels are properly assigned. How long did you leave Auto-RF turned on? Or are just some areas affected by frequent changes?

In that case, i guess Rob is right and the best advice is to turn off Auto-RF after an intitial DCA (or at least override the settings for the problematic APs).

Another thing i've found is the CLI command "config advanced 802.11b tx-power-control-thresh [threshold]" on the WLC. With this command you can change the threshold in dBm the WLC assumes to adjust power settings. Decreasing this threshold could also decrease channel changes, but it could bear some other problems (Co-Channel interference, inadequate coverage for VoWLAN e.g.). Just a quick thought.

Greets,

Sebastian