Wireless Access

I'm tweaking CSR values in a high density deployment and am trying to find the best way to measure the  effects on an AP.  The easiest way I figured was to check an AP's neighbor signal strength before and after.  For example: NF is -97dBm and a neighboring AP is seen at -90dBm.  I put in a CSR of 10, so the AP should tune out anything higher than -87dBm.  However, the AP can still see the neighbor at -90.  I waited up to 10 minutes to make sure ARM had plenty of time to take measurements, but had the same result.  Unless ARM takes all of its calculations into affect, sans-CSR, I don't understand why I'm not getting the expected result.  Any thoughts or other ways of measuring the CSR effects?

Thecompnerd,

Implementing CSR is a black art and very rarely done. The reason why you can still see your access point, is that the client can still see the beacons... The AP probably will not answer the probe responses, but clients can still see the beacons is why it is such a black Art and should not be used with clientmatch.

Sorry, I had to correct my wording above.  I did not mean to say that "I" could still see the AP (of course I could)...  I meant that the AP could still see its neighbor at -90 after making the CSR adjustment.  I do not expect the AP to see its neighbor in this case.

That being said, I understand the feature and pitfalls of using it.  Just can't for the life of me figure out a useful way of measuring its affect on other APs.

Hey Jay,

 

I don't know if this applies, one of the Dark Wizards may be able to answer, but the APs may not apply the same CSR rules to other APs since the From/To DS bits are different than a client. The true test would be to see, via a pcap more than likely (PITA I know), if the AP ceases to respond to client probe responses below your CSR threshold. Maybe set this up on a test AP with a test SSID, configure a client for said SSID, associate/authenticate, start your PCAP (if you have a MAC you can do this natively, otherwise you can do the pcap on the controller and pull off via the logs, or use an air sniffer which I am SURE you have), then walk away slowly making note of the SNR of the frames in the capture. As you get close to the threshold, you should start to see the frames stop. Maybe make Ryan help you here and he can carry the test device while you follow on the capture.

There may also be some commands that can tease out the unobservable effects, but honestly, CSR is kind of a sledgehammer in regards to the radio (very much NOT a scalpel), so with respect to client-thresholding, chances are any frames below the threshold are getting dropped in the bit bucket without much of a trace.

 

Btw, I am no longer in Fed, as I moved to be the Technical Marketing Engineer for Outdoor, APs, Accessories, and Federal (weird hodge-podge). So if/when you start your outdoor work, let me know. 

Whoops!  Somehow missed your reply, Jerrod.  Decided I needed to update this thread after doing some testing in the lab.

 

I was looking for an easy way out since I was trying to test CSR after-hours, remotely, and the main concern has been interference from CCI.  Although, I will agree that your test methodology is much more complete and ultimately what I should do.

 

I did some testing in my home lab and noticed that CSR did not have an immediate effect, just like I noted in my original post.  However, if you reboot the AP, CSR seems to take effect.  The only confirmation I have of this is by testing the output of 'show ap arm neighbors' before/after adding a CSR value and rebooting the AP.  After the reboot, several APs below the adjusted receive sensitivity disappeared.

 

Also, I noticed any associated clients would start having low levels of packet loss immediately after adding the CSR value, but before rebooting.  I would've expected 100% packet loss given a CSR of '55'; obviously extreme, but I was trying to get quick results.  If I deauthenticated the client, it was then unable to associate to the AP until I got within range.  I was rushing this a bit, so I didn't setup a WLAN capture, but I'm sure I wouldn't have seen probe responses.  I will try this again tomorrow.

 

Doh!  Could've used your help with some of the outdoor APs we just implemented.

 

EDIT

 

On second thought, in the case of the neighboring APs, perhaps ARM is caching neighbor entries for a period of time so it only appears as though the AP is still able to hear neighbors after entering a CSR value.  Rebooting would automatically clear this information.

thecompnerd,

 

The neighbor APs information is cached 30 minutes after we last heard from them, so that would explain why you are stil seeing it after you made the change.  Rebooting the AP should purge entries heard by that AP.

(Aruba7640-US) #show wms general | include ap-ageout
ap-ageout-interval            30

The max CSR number processed for the AP225 is about 25, so anything above that would probably have no effect.