Cisco’s CleanAir Spectrum Analysis: Review

Air quality

The second element, Air Quality, essentially aggregates the severity index of all identified interferers to provide a score to help qualify how usable the AP’s RF environment is.

The WLC aggregates these information elements from all attached CleanAir APs, presenting per-band views on the average and minimum Air Quality assessments for the channels scanned by AP. Air Quality scores update every 15 minutes, unless an admin is looking at a specific AP and radio, which bumps the update to every 30 seconds, providing relatively real-time assessments.

The channels scanned by an access point depend on its mode.  APs in local mode (servicing wireless clients) only monitor the RF for the channel that the AP transmits on, while monitor-only mode APs constantly cycle through all channels, passively listening to both Wi-Fi and non-Wi-Fi traffic.

In my tests, CleanAir identified an assortment of the usual suspects of interferers, providing impact assessments for each: Bluetooth links and discovery actions, dozens of DECT phones in the 2.4 and 5 GHz bands, and microwave ovens. CleanAir also classified some X-box controllers, analogue wireless cameras and generic time-division duplex transmitters.

Because interferers could be detected by several CleanAir APs listening on the same channel, the WLC attempts to merge together interferer records to avoid multiple listings for the same interferer. For local-mode APs, neighbour lists are used to determine if detecting APs are near each other. The WLC tries to cluster records it thinks are for the same interferer into one object, with the AP most affected by the interferer as the cluster centre.  Merging is also done at the MSE for larger networks supporting multiple controllers, and the MSE is also required to merge records provided by a monitor-mode-only AP.

I found merging to be somewhat hit and miss, not unexpected given the newness of the capability.  Although frequently able to correctly merge my interference sources, CleanAir left numerous like items orphaned. It also merged a few interferers incorrectly, such as the time an Apple Magic Pad that had never before entered the building was merged with a two-week-old otherwise unknown Bluetooth source.

Interference avoidance

CleanAir also helps automate interference avoidance, as CleanAir is integrated into Cisco’s RRM (Radio Resource Management) feature set, thereby allowing the network to change an AP’s channel if Air Quality drops too low for too long due to either Wi-Fi and non-Wi-Fi causes.

CleanAir provides two types of interference avoidance for RF sources. An event-driven RRM is utilised for intermittent interference, as it detects interference overpowering AP and moves the AP to a different channel, providing a cleaner Air Quality for at least three hours–even if the new channel is the same as that of a neighboring AP.

Persistent Device Avoidance, meanwhile, tracks interferers that tend to be always on and rather immobile–such as wireless cameras or Canopy outdoor networking equipment. If CleanAir classifies the interferer as persistent, the channel change will last until seven days have passed since the last time the interferer was detected.

Changing channels is tricky business in a large network, and Cisco put the time thresholds in place to avoid APs constantly thrashing between channels, which has implications for the performance of other parts of the network. And CleanAir RRM makes its changes with some smarts, avoiding channel changes due to interference from frequency-hopping devices that affect many channels by bouncing their transmissions around the spectrum.

In tests, I found CleanAir RRM worked quite well, quickly identifying and avoiding interference provided by a pair of incredibly noisy wireless cameras–one using the 2.4 GHz band, the other 5 GHz. In practice, I have some qualms about using event-driven RRM in the 2.4-GHz band, however; there simply aren’t many alternate channels to choose from. For instance, I set the 2.4-GHz camera on a channel that affects Wi-Fi channels 1-9, so four of my six APs immediately flopped over to channel 11, where they stayed for the next three hours.

Since the other channels were swamped with a high-power, high-duty cycle interferer, the alternative was not ideal either, but four adjacent APs on the same channel in a multicell architecture is a practice that’s typically frowned upon. At the very least, I’d recommend tuning the non-Wi-Fi RRM thresholds very low, say, in the 2.4 GHz band.

Thankfully, CleanAir trap and action thresholds are highly configurable, which allowed me to tailor alarm and RRM actions differently for each band or for each controller. Typically, many interferers—particularly Bluetooth devices—are relatively low risk for network performance, so I could also configure the system to avoid reporting IDRs for those devices, while still accounting for their presence in the Air Quality score.