How Do I Use Auto AP Capacity Management ?
Preseem Auto AP Capacity Management enables operators to maintain a high quality of experience (QoE) when an Access Point is overloaded.
Capacity Management Overview
When network elements like Access Points reach their capacity, the subscriber quality of experience (QoE) degrades significantly. This occurs because many devices implement buffering with simple FIFO queues, which increase latency and eventually packet loss when the outgoing link reaches capacity. Preseem's capacity management uses modern active queue management (AQM) techniques that do not have this behavior and deliver a great subscriber experience even when the link is busy. Preseem's traffic management can be deployed on one or both of the subscriber plan and the Access Point levels.
On the subscriber level, we refer to this as QoE-optimized plan enforcement. This feature ensures that a subscriber continues to have a good experience even when fully using their plan rate.. Concretely, this solves problems like games or VoIP calls not working while other members of the household are watching streaming video or downloading large files.
The most common bottleneck in the network is a subscriber hitting the limit of their plan. Another typical bottleneck is the Access Point itself. Preseem's Access Point Capacity Management enables the operator to set rate allocations for the Access Point as a whole and (within that) apply intelligent AQM algorithms and share the traffic between subscribers. The allocated rates can be configured directly in Preseem when viewing an individual Access Point.
Auto AP Capacity Management strives to automatically balance throughput and latency as a significant improvement over manual management. As this might not meet specific business goals on every AP, you can control how management is applied and enable or disable Auto or manual management for individual APs at will.
Benefits
All wireless connections have some inherent, or baseline, latency. This latency varies based on factors like the specific wireless framing techniques used (Wi-Fi, LTE, and proprietary PHY will all vary), and other factors like the frame length and frame scheduling. When an AP is idle, most packets traversing the link should experience the baseline latency. As an AP becomes busier, when packets arrive from the backhaul at a faster rate than the wireless link(s) can transmit, this can cause either packet loss or packet buffering, which increases latency over baseline.
Enabling an appropriate allocation of managed rates on an AP will not dramatically affect QoE or reduce measured latency when an access point is not under load. However, even a nearly idle AP will receive some benefit, as traffic downstream from the Internet can have microbursts of traffic that exceed the AP's link rate. Managing those microbursts does help the affected packets, but may not be a large percentage of traffic overall; the improvements may not be easily visible when measured.
At the other end of the range, when an Access Point has < 25% airtime free, we find dramatic improvements in the 95th percentile of latency — in the range of 20 to 40 ms of lowered latency — when allocations are properly calculated and applied at the Access Point level. When the AP is busy, there will be many packets to measure.
When looking at the 80th percentile of latency, we find that manually managed rates are not usually recommended — AP link rates shift over time, and it becomes an administrative burden to keep them optimized. Auto management produces an improvement of around 5 ms in the 80th percentile, a statistically significant result. Framing mode (flex vs. fixed) will affect the results, with flex (where available) having lowest baseline latency as well as lowest overall latency.
Requirements
- Operate permissions on your user and Preseem version >= 1.6
- All subscriber traffic traversing the AP must be correctly mapped to network topology
- For Auto mode, the specific AP model must have Auto support
- For best results, ensure stable and consistent RF link rates — fluctuating link rates reduce the effectiveness
Explore interface
Network Interface
Auto AP Capacity Management
Auto AP Capacity Management
Auto Overview
Automatic rates are calculated at least daily for all supported models. If AP Capacity Management and Auto are both enabled on a supported AP, the calculated rates are also applied to the active managed rates. Metric charts in the Performance tab show over time when allocations are used (Tx and Rx direction) and the resulting managed rates.
Usage
For a single AP, checking or unchecking the Auto AP Capacity Management box under Access Point -> Config is sufficient to enable or disable full Auto.
If a model is not yet supported, Auto cannot be enabled, but manually managed rates may still be used.
To enable auto allocations as new APs come online or new models become supported, toggle Auto Capacity Management For New APs, available in the hamburger menu at Network -> Access Points -> Capacity, to "On". This sets a company-wide default:
To make a global change to all currently supported APs, use 'Enable All Auto Capacity Management' or 'Disable All Auto Capacity Management'.
To view all allocations, use the list at Network -> Access Points -> Capacity, which contains all allocations or managed rates, and any errors or other informational alerts.
Errors and Notifications
There are several classes of errors and notifications possible on a given managed entity:
- Unsupported
- This model is currently unsupported for auto capacity but may be manually managed.
- Tx/Rx Low
- This AP exhibits unusually poor RF link rates, under 2 bits per herz in efficiency. (Some very old equipment may exhibit this under normal conditions.) Producing good QoE results by managing APs at the lower end of the efficiency scale is challenging, so while Preseem can still calculate auto AP rates, we apply them less aggressively. We recommend improving and stabilizing the link rates on clients connected to these APs for the best business value. Advanced Fixed Wireless BV and RF scores can help with this process.
- Low Samples
- This AP was not pollable and online at the current channel width for at least 80% of the previous day; the most recent calculated updates skipped this AP and made no change. This error will clear (and rates will be updated) when the channel width is stable, and enough data is collected.
- Flex Framing (warning)
- Auto AP capacity management can work with flex framing, but requires careful consideration of your business goals. While previously considered problematic, we have customers successfully using AP capacity management with flex framing when their primary goal is lower baseline latency rather than maximum throughput. Flex framing capacity planning is more complex than fixed. See more details on flex framing below.
FAQ
- What are the prerequisites for using capacity management?
- Access Point bandwidth allocations require Preseem 1.6 or greater (preferably the newest release) configured with the TwoLevelMQ policy. This has been the default for new installs for some time. Preseem support can verify that your node is ready for Access Point capacity management if you have any concerns.
- What are the requirements for Auto?
- An AP model must be supported for capacity management, and the AP must not be in an Error state in the AP list — common errors might be a lack of SNMP or HTTPS credentials. In general, 80% of daily KPI records at a given channel width must be present for Auto rates to calculate.
- What should I do when I change RF channel width or framing mode, and allocations are on? Won't the link rates no longer match the allocated rates?
- You have two choices.
- 1) Set manual override rates that match the new link rates (or enter 0 or blank to turn management effectively off). If Auto is on, leave it on.
- or 2) Leave the existing rates as they are. If Auto is on, the newly calculated rates will take effect on the second day. (A full day's worth of samples is required to calculate new rates.)
- You have two choices.
Flex Framing
Understanding Flex Framing with AP Capacity Management
For any form of management to reduce latency, Preseem must be the bottleneck. When bottlenecks occur elsewhere, rises in latency are uncontrolled. With flexible framing, the airtime available can be an unconstrained bottleneck, particularly in the upstream direction, which can make latency control challenging.
When you add allocations to a flex framing AP
- You create a control point in the downstream direction
- Upstream doesn't get a true bottleneck, though the rate management still helps
- The AP essentially operates like virtual fixed framing, but without synchronized clocks with other APs
Preseem implements a sophisticated algorithm that detects the desired UL/DL demand ratio and optimizes this dynamically over time
When to Consider Fixed Framing with AP Capacity Management
- You must have frequency reuse on the tower. This results in highest overall system capacity, considering a wide geographical area. This configuration is most common with Cambium APs.
- For Ubiquiti Airmax AC:
- You have 40 MHz or larger channel sizes and fewer than 20-30 clients per AP.
- You have fewer than 20 clients, are on 20 MHz channels, and are using a 67/33 split with 8 ms frame sizes. Other configurations display upload ACK starvation causing download performance issues.
When to Consider Flex Framing with AP Capacity Management
- Your primary goal is the lowest possible baseline latency, but accept variable latency
- You're willing to trade some overall system capacity for better individual CPE performance
- You want to avoid UL ACK starvation issues that can occur with fixed framing at aggressive splits (e.g., 75/25)
- You want to use airtime most efficiently because the UL/DL split can shift depending on need.
As one example, using AP allocations with AirmaxAC Flexible framing allows you to keep an AP on 20MHz with an effective 75/25 DL/UL split but avoid the UL ACK starvation issues common with 75/25 fixed-framing. You're directly allocating the bandwidth, but not the airtime (or frame size) that can be allocated to the UL direction. This is different from fixed-framing which constrains airtime usage at the frame level where only so much data can fit into that 25% of frame time.
Trade-offs Summary
- Fixed Framing with AP Capacity Management: Best for overall system capacity and predictability, with slightly higher baseline latency
- Flex Framing with AP Capacity Management: Best for the lowest baseline latency at the expense of overall system efficiency
While Preseem's algorithm is adaptive to user demand, we recommend running a trial to ensure this configuration meets your specific business objectives.
Further Reading: Fixed vs. Flex Framing
Why Fixed Framing is Generally Preferred
In a system where multiple APs can hear each other, fixed frame scheduling allows more frequency reuse — APs transmit and receive simultaneously. This leaves potentially more airtime silent but allows more total receivers at a given moment. The system as a whole achieves more aggregate capacity. Flex framing allows an individual CPE to have more capacity at the expense of overall system efficiency.
Detailed Comparison
Compare... |
Flex Framing |
Fixed Framing |
Baseline Latency |
Lower baseline latency. A CPE does not need to wait as long for an airtime slot to send a response back. |
Slightly higher baseline latency. A CPE may need to wait for an airtime slot to transmit a response. |
Loaded Latency |
Higher latency variability (jitter). Worse for VOIP. Similar peak latency to fixed framing. |
Lower latency variability (jitter). Better for VOIP. |
Link Rates |
Lower total system capacity considering towers and regions. Higher, half-duplex link rates. Maximum CPE throughput in a single direction is higher. |
Overall improved system capacity considering towers and regions. Lower, full-duplex link rates. Maximum CPE throughput in a single direction is lower. |
Downstream Airtime |
Without management, downstream can instantaneously consume all airtime, strangling upstream. Capacity is unpredictable. |
Downstream is constrained and cannot restrict upstream. Capacity is predictable. |
Upstream Airtime |
Without management, upstream can instantaneously consume all airtime, strangling downstream. Capacity is unpredictable. |
Upstream is constrained and cannot restrict downstream. Capacity is predictable. |
Capacity Planning |
More difficult capacity planning as airtime constraints are dictated by demand. |
Easier capacity planning as airtime constraints are fixed. |
Plan Speeds |
Plan Speeds are sometimes oversold. For example, selling a 100/100 Mbps plan on a link speed of 100 Mbps means that downstream and upstream speeds can only be achieved sequentially, and a simultaneous DL/UL speed test fails. |
Most framing modes are asymmetric (more down than up). Plans sold are typically asymmetric. Downstream and upstream are independent, and a simultaneous DL/UL speed test should succeed. |
RF Efficiency |
No special frequency re-use. Transmitters can interfere if not adequately separated spatially and by frequency. |
Some frequency re-use. Transmitters can co-exist with proper spatial and frequency planning. |
Manual AP Capacity Management
We no longer recommend manually managing rates. The accuracy of manual capacity management has a short half-life — link rates tend to fluctuate. One potential use is to artificially cap site bandwidth below the capacity of a backhaul. With that in mind, when first enabling Access Point capacity management, Preseem recommends these approaches:
- If the Access Point is congested, set the rates to 10% below the observed peak rates. This should bring the capacity below the pain threshold measured by latency and other metrics such as TCP retransmissions. If no improvement is observed, further reduce the managed rates by another 10%.
- If the Access Point is not currently congested, the Access Point comparisons (as displayed by the orange line on the Experience charts) provide the best starting point. The Access Point comparisons are built from data collected from hundreds of customers and tens of thousands of Access Points and give a robust indication of what the Access Point should be capable of in the real world (vs. marketing numbers). For more real-world Access Point performance information, see our Fixed Wireless Report.