AI-RRM
Click 日本語 for Japanese
Learn more with these free online training courses on the Meraki Learning Hub:
Overview
Radio resource management (RRM), traditionally known as Auto RF for the Meraki cloud network, is a solution that improves wireless networks by automatically optimizing radio configurations (such as channel changes, transmit power, and more) to mitigate interference and maximize RF spectrum capacity. Adding artificial intelligence (AI) to RRM offloads analysis of RF patterns to an AI engine that more intelligently optimizes RRM. The result is better connectivity for users and simplified management for network administrators. Unlike traditional RF optimization systems, which use snapshot data as the source of decision-making, AI-RRM uses historical, long-term, trend-based RF data lakes to automate and optimize RF. Every set of traditional RF optimization decisions (such as channel decision spectrum allocation and AP power levels) now gets AI/ML-based decisions.
Cisco AI-RRM is different as it allows the analysis of historical dynamic RF data over time. Evaluating complex RF data often results from factoring in local knowledge of “Normal” against the currently displayed data. “Normal” can, and does, vary from site to site based on equipment choices and architectural design vs. client density.
Together, the AI-RRM algorithms with the Cisco AI Cloud take Wi-Fi RF management to a new level that correlates 24x7 observations from the network and client devices and applies 20+ years of Cisco RF technology to drive user experiences into the future.
Benefits of AI-RRM
Improved Wireless Connectivity (User Experience)
AI algorithms enhance wireless network performance by proactively learning your network's trends and patterns, improving the operation of wireless endpoints over time. This intelligent optimization leads to significant improvements, including a 40% reduction in co-channel interference and a 7-decibel gain in signal-to-noise ratio for wireless clients. Additionally, the system includes a "Busy hour" feature that reduces RRM changes during peak times by up to 99%, ensuring 24/7 connectivity.
Lower Operational Costs (Administrative Experience)
AI simplifies RRM and provides a scalable deployment workflow, even for users with minimal RF expertise. It offers RF visibility and actionable insights to optimize RRM profiles, increasing efficiency and eliminating error-prone manual processes.
Maximized Investment
Whether used with Wi-Fi 6E, Wi-Fi 6, or previous generations of Wi-Fi at any scale, AI-RRM offers visibility into previously unknown aspects of your network.
To learn more about the benefits of AI-RRM, visit: What is AI-Enhanced Radio Resource Management?
The AI-RRM topology flowchart steps are recorded here:
- Anonymized RF data is sent from the access point to the AI Cloud which utilizes AI-Enhanced RRM algorithms.
- After data analysis, RF optimizations are then automatically applied to the access point.
- Wireless endpoints are able to benefit from this optimized wireless experience.
How to enable AI-RRM
The new RRM tab is introduced on the Radio settings page. It consists of all the AI-powered RRM features that majorly drive decisions based on historical data, along with many Cisco Radio Resource Management features that will help optimize the RF network much more efficiently and sophisticatedly.
AI-RRM requires firmware version 30.x or newer and MR-ADV license.
AI-RRM is available as an Early Access feature in the US and EU regions. AI-RRM is not supported in APAC and Canada during Early Access, but will be supported in these regions when the feature becomes generally available.
Navigate to Wireless > Configure > Radio Settings > RRM tab > Enable AI-RRM, as seen in the image below.
Make sure to enable AI-RRM during a maintenance window.
AI-RRM will utilize the RF parameters specified in each RF profile to implement changes in RF parameters. Across the network, access points may either share the same RF profile or be assigned distinct ones. AI-RRM does not modify or change your current RF Profile settings or assignment.
Refer to the following documentation on how to create an RF profile.
Refer to Meraki Wireless for Enterprise Best Practices- RF Design.
Flexible Radio Assignment
In a dense AP deployment, there may be more 2.4 GHz radios than necessary for the coverage area. This usually causes high co-channel interference with the limited number of channels available on the 2.4 GHz spectrum.
The Flexible Radio Assignment (FRA) functionality in AI-RRM enables the dynamic deactivation of 2.4 GHz Tx radios in the AP. The radios are deactivated when they are acting as a source of co-channel interference, are adding to radio overhead, or are not required to maintain optimal coverage and roaming experiences. This feature helps to minimize interference among redundant neighboring 2.4 GHz cells.
This feature will help reduce the co-channel interference in the 2.4 GHz environment, ensuring clients have adequate, full coverage without manually changing transmit power configurations or redesigning your Wi-Fi architecture.
This feature can be enabled by toggling the enable button under Flexible Radio Assignment.
The Flexible Radio Assignment feature can only be used when AI-RRM is enabled because FRA decisions are based on trend-based AI/ML decisions.
How to identify which 2.4 GHz radios are disabled?
In the Radio settings Overview page, if you select 2.4 GHz in band, you will find APs with Target Power as Off.
If FRA is disabled, the 2.4 GHz radios that were turned off by FRA will automatically be re-enabled within 30 minutes.
AI Channel Planning
A well-designed channel plan will ensure we get the most out of the RF network to operate as a high-performance wireless network. To achieve this, use a wide spread of channel availability in the AutoChannel list. However, these channels can get affected by DFS events and non-Wi-Fi interferers, such as jammers, which reduces the number of channels that can effectively be used for wireless network operations. AI channel planning maintains the channel avoid list on a per-access point basis in the network.
AI channel planning will avoid channels that are most affected by the following reasons:
- Too frequent DFS events
- Too frequent jammed channel
- AFC low power regulation (outdoor 6 GHz AP)
There are 2 aspects to AI channel planning; RF Jammed and DFS Hit events.
Once a DFS hit or jam event occurs, the channel will be put on a Channel Avoid List.
- If a channel is in avoid mode, we strictly avoid using this channel in Auto RF unless all the channels are avoided.
- The avoid period for this channel to stay in the Channel Avoid List depends on the severity.
For channels in avoid mode, the table below indicates the monitoring period for each level of severity:
|
Severity |
Monitor Period |
|---|---|
|
6 |
1 day |
|
5 |
2 days |
|
4 |
3 days |
|
3 |
4 days |
|
2 |
5 days |
|
1 |
6 days |
The figure below shows the insights and learnings for channel avoidance that can be exported via CSV.
Download historical report will contain a CSV with AI Channel Planning decisions and a list of channels affected, including the start and end time for avoidance.
Busy Hour
Reliable connectivity is key when it comes to an efficient wireless network. This is essential during the active hours of any business vertical. The Busy hour feature will avoid unnecessary RF changes during the defined period and make the essential changes after the busy hour. This will eliminate client disconnections or roaming due to channel changes.
Each business vertical has its own characteristic of busy hours, and sometimes it is hard to dictate from the central network admin perspective. With the power of the dashboard, Auto RF can now optimize each network individually for its unique daily busy hour.
During the Busy hour, Auto RF will only change an AP's radio transmit power if required and minimize the AutoChannel calculation. Some exceptions will be DFS, jammed channels, access point addition or the Update AutoChannel button from the dashboard.
Crucial changes include:
- DFS hits
- RF Jammed
- Access point addition
- Persistent, excessive co-channel interferences from nearby rogues
- Update AutoChannel button from the dashboard
Sensitivity outside the configured busy hour is equivalent to high sensitivity.
As shown in the figure below, Busy hour assignments can be set in 2 ways:
1. Manual: Define the network’s active hours.
2. Auto: Let the dashboard recommend the busy hour. Our system will analyze your past six weeks of data to understand your active hours and set it automatically. The auto Busy hour is calculated based on the number of clients connected to the network and the amount of traffic sent/received.
Make sure to have your Network Time zone configured accurately.
Radio Settings Overview Page
Below are the headlines regarding the RF Coverage and RF Performance, which includes the overall RF Performance score (100-0, with 100 being excellent) as well as highlights such as % APs with poor co-channel interference and the number of Mitigations done by AI channel planning. The RF Coverage summary looks at the AP density (the number of AP neighbors seen at or above -70 dBm) and Connectivity – the average client SNR during the last RRM run.
RF Dashboard
The summary shows the total number of online access points and Clients in the specific filter selection.
Summary
The Summary shows the total number of access points with the radios enabled with the selected band filter, along with connected clients in these radios. This shows the overall AP and client count in the entire network and does not separate it depending on different RF Profile assignments.
RF Performance
RF Health is a metric used to measure your RF parameters, which can lead to a good or bad Wi-Fi experience. RF Health in the Hero bar averages all the APs selected in the above filter.
How is RF Health calculated?
It is a weighted average of the following factors:
- Co-channel Neighbor Utilization:
This process compiles a list of APs operating on the same channel as the current AP and assigns a weight to each based on their channel utilization and proximity (signal strength). It correlates the activity of nearby APs with their impact on the current AP. The impact of each AP is cumulative. The goal is to show that closer APs with higher channel utilization (higher RSSI) have a larger impact on RF health. - Co-channel Overlapping:
It involves compiling a list of nearby access points (APs) on the same channel and correlating their operating power (Transmit Power Control - TPC) with their RF distance (signal strength). It evaluates how much these nearby APs overlap with the evaluated AP's operating channel. The goal is to show that closer APs with higher operating power (higher RSSI) have a larger impact on RF health, regardless of their current transmission utilization. The impact is cumulative for all APs on the same channel as the evaluated AP. - Noise Side Channel:
This metric correlates the impact of detected noise with the current operating channel and the "channel distance" where the noise was detected. - Noise Same Channel:
The procedure involves converting noise measurements into a 0 to 100 scale, called compensated noise. Noise levels below -80 dBm have 0 impact, while noise levels above -50 dBm are considered to have 100% impact. This conversion is straightforward, without any side channel subtraction, directly translating received noise power levels to the 0-100 scale based on these parameters. - Co-channel interference:
The passage discusses the impact of Wi-Fi activity on the channel, similar to noise correlation but specific to Wi-Fi interference. Access points can coexist better with Wi-Fi interference than random noise. The impact range is defined, where a value of -50 indicates full impact (100%) and -90 indicates no impact (0%). In Radio Resource Management (RRM) metrics, interference is measured in terms of "time" percentage. Any value exceeding 30% time is considered to have a full impact (100%). - Adjacent channel interference:
This metric is similar to noise correlation but focuses on interference. APs typically manage interference (Wi-Fi activity) better than random noise. A value of -50 indicates a full impact (100%), while -90 indicates no impact (0%). In RRM metrics, interference is measured as a time percentage, with anything over 30% time considered full impact (100%). - Low SNR Clients:
The objective is to convert client connections with poor SNR levels (<=20dBm) to a 0 to 100 scale. APs with consistently high numbers of low SNR clients may indicate radio issues on nearby APs, coverage problems (poor deployment), or client roaming issues (sticky clients). This evaluation excludes APs with fewer than 5 clients. - Radio Utilization:
This view differentiates between "good" and "bad" channel utilization. "Good" utilization refers to the AP's transmissions and receptions, while "bad" utilization indicates external energy on the channel that does not involve the AP. The goal is to maximize "good" utilization, as it reflects the AP performing its intended functions. This metric will highlight instances where non-AP or external utilization exceeds AP activity, indicating potential interference. - Channel Change Health:
This metric tracks channel changes relative to AP uptime in days and flags APs experiencing more than four-channel changes per day, as this could negatively affect client performance.
The table below shows individual AP RF Health scores listed under the RF Score column.
High CCI %
The High CCI value measures how many of your wireless access points are experiencing a lot of co-channel interference. Interference can make your Wi-Fi slow or unreliable.
To calculate this, you:
1. Count the total number of APs that have poor radio performance (these are the ones struggling with interference).
2. Multiply that number by 100 to turn it into a percentage.
3. Divide by the total number of all APs you have.
This percentage tells you how big of a problem interference is in your network. If the number is high, a lot of your APs are struggling with interference.
Mitigations
The Mitigations events showcase the total number of entries in the channel avoid list table related to the APs in the network, as identified by AI channel planning. This information directly correlates with the downloadable CSV file available in the AI channel planning section on this page.
RF Coverage
AP density shows how many neighbors each AP has, on average, in the network. It uses the Neighbor Density Metric (NDM) to calculate AP density.
The calculation for the NDM is as follows:
NDM is determined by counting the number of 'strong' neighbors, defined as neighboring devices with a Received Signal Strength Indicator (RSSI) greater than -70.
The categorization of Radio Frequency (RF) density is based on the average NDM and is classified in the following manner:
● Very High: The average NDM is greater than 15.
● High: The average NDM is greater than 10 but less than or equal to 15.
● Medium: The average NDM is greater than 5 but less than or equal to 10.
● Low: The average NDM is less than or equal to 5.
FAQs
-
During the Busy hour, what will happen if a manual change is made on an AP?
-
All the APs in the network will operate as usual unless there is a significant change in the channel assignment of the AP. If there are significant enough differences between the new candidate channel vs. the current operational channel, it will change.
-
- If changes are made to a RF Profile during the Busy hour, does it wait until after the Busy hour?
- New additions will not reset the busy hour criteria, which is much more sticky than a non-busy hour.
- Will auto Busy hour work with templates?
- No, auto Busy hour is not supported on templates.
- What if a customer has all of their assets under the same network spanning across multiple locations and timezones?
- Auto Busy hour can calculate one busy hour fitting for all the APs in a particular network. Different locations need to be in different networks with appropriate timezone configurations.
- How do you track the severity table of the network, and is it per AP?
- It is per AP, per channel, and per event type.
- What are the configuration options that are being modified to facilitate the channel avoidance?
- No configuration is needed on the network. The algorithm will monitor the number of DFS hits and decide the severity of the channel.
- Is there an overview of the scheduling the channel changes?
- One can view the overview in the Auto RF event log.
- What metrics are used to determine the Busy hour - is it combination of current client count, usage (effectively what is displayed on the summary report)?
- Combination of client count and usage.