Cluster Optimization vs. Auto Channel

Auto channel: why is it not a "silver bullet"?

The Auto-channel means that Wi-Fi router selects channels automatically based on the vendor’s propriate algorithms. In this case, a device takes into consideration only itself, that can increase the whole system interference! Practically, additional optimization algorithms are required to make the optimization process stable and manageable.

Most APs only select a channel at boot time, but a channel that was good enough when AP was rebooted last time can be a poor choice after a few days, weeks, or months.

The distribution of channel changes.
The distribution of channel changes

Most people don't want to delay booting by spending enough time to truly evaluate each channel, so they use poor heuristics like "just pick the channel where you can see the fewest number of APs", which doesn't necessarily correlate to which channel will provide the best throughput and reliability.

To sum up:

  • AP Auto-channel functionality is not an efficient method of management and engineering of wireless networks, especially when the networks are heavily congested.
  • Centralized network Radio Resource Management is essential in congested wireless networks ecosystems.
  • Efficient wireless networks Radio Resource Management (RRM) solutions must be capable of treating devices in their network neighborhood accordingly and adaptively to their class (Managed AP or Alien AP classes).

Cluster optimization algorithms

Sometimes optimization of the single device may decrease overall performance of the wireless network. Clusters based optimization is proposed to avoid these situations.

Example of cluster optimization: radars (devices under the control) are green, aliens (devices out of control) are red
Example of cluster optimization: radars (devices under the control) are green, aliens (devices out of control) are red

However, what are criteria for selecting the best channels? Channel quality weights (CQW) are the quality metrics for all RF channels in a band.

CQW’s are calculated by RRM (Radio resource management) from collected data from access point. The channel is assigned to devices for broadcasting based on the maximum available CQW within the predefined user and country restrictions. RRM algorithms take into account devices controlled by the same provider (Managed AP devices) and organize a wireless network for minimal managed AP interference.

TODO
RRM receives Radio information from the CPE APs, runs algorithms in order to estimate the current environment, and finally makes changes if needed in the APs configuration. RRM has a REST interface and is integrated with device management systems (DMS), for example: ACS(CWMP TR-069), Controller(USP TR-369) or any others).

Device throughput is usually affected by interference from other Wi-Fi devices in the network environment, as well as radio signals/noise in the spectrum other than Wi-Fi. Wi-Fi interference is created by two components: interference with Alien AP (not controlled by ISP) devices and Managed AP (ISP controlled) devices. They are technically similar, but algorithms can handle them differently, minimizing the effects for Managed AP devices.

When the CPE is set to another channel, the CQW of the initial channel gets higher, while the CQW of the target channel degrades. CQW strongly correlates with the overall interference seen by CPE.

Channel Selection Algorithms should include comparison of the initial channel CQW (which is a reference point) and the target channel CQW with the assumption that AP switched to that channel.

The most obvious implementation of the cluster optimization is the “brute force” way. Let us call it "Brute Force algorithm". The main idea of the "Brute Force algorithm" is to select the best channel combination from all the possible combinations. Complexity of this algorithm is O(N^m) where N is the number of radars in the cluster, and m is the maximum number of the possible channels. The block-scheme of this algorithm is presented below.

The block-scheme of the "Brute Force" cluster optimization algorithm.
The block-scheme of the "Brute Force" cluster optimization algorithm

This implementation is optimal in terms of accuracy, however, is unacceptable for large clusters due to complexity. Therefore we use several heuristics to achieve optimal balance between quality and complexity.

The parameters required for calculation of the CQWs values can be obtained, for example, via TR-069 or TR-369 protocols from the following data fields:

Device.WiFi.Radio.{i}.

(information about Wi-Fi radar)

NameTypeDescription
ChannelunsignedInt­[1:255]

The current radio channel used by the connection. To request automatic channel selection, set AutoChannelEnable to true.

Whenever AutoChannelEnable is true, the value of the Channel parameter MUST be the channel selected by the automatic channel selection procedure.

Note: Valid Channel values depend on the OperatingFrequencyBand and RegulatoryDomain values specified.

CurrentOperatingChannelBandwidthstringThe channel bandwidth currently in use. Enumeration of:
  • 20MHz
  • 40MHz
  • 80MHz
  • 160MHz
PossibleChannelsstring­(1024)

Comma-separated list (maximum list length 1024) of strings. List items represent possible radio channels for the wireless standard (a, b, g, n) and the regulatory domain.

Ranges in the form "n-m" are permitted.

For example, for 802.11b and North America, would be "1-11".

Device.WiFi.NeighboringWiFiDiagnostic.

(information about neighbors of the Wi-Fi radar)

NameTypeDescription
SSIDstring­(32)The current service set identifier in use by the neighboring WiFi SSID. The value MAY be empty for hidden SSIDs.
BSSIDstring­(17)[MACAddress] The BSSID used for the neighboring WiFi SSID.
ChannelunsignedInt­[1:255]The current radio channel used by the neighboring WiFi radio.
SignalStrengthint­[-200:0]An indicator of radio signal strength (RSSI) of the neighboring WiFi radio measured in dBm, as an average of the last 100 packets received.
OperatingChannelBandwidthstringIndicates the bandwidth at which the channel is operating. Enumeration of:
  • 20MHz
  • 40MHz
  • 80MHz
  • 160MHz
  • Auto

Besides the tasks on optimization of interference, our solution for implementation of RRM allows you to analyze the operation of a Wi-Fi network. For instance:

  • To analyze the current state of the signal levels of associated devices. If the signal strength between the router and device is too low, this may indicate the need for planning of the router’s location.
  • To analyze the current statistics on competitors: the types of used encryption, the used terminal equipment and WiFi standards, etc.
  • To get statistics for specific devices.

The solution that we offer can be either integrated with the customer's DMS systems or delivered together with our DMS system. The solution can be deployed either locally or using cloud technologies, depending on the preferences of the customers.