Bandwidth planning and capacity calculations

One of the key success criteria of a high performance, Wi-Fi network is the expected system performance and amount of bandwidth available to users and applications.

Number of access points

Once considered, a simple formula can help you estimate the number of access points required for a high-density deployment:

$AP\ count = 5/6\ GHz\ radio\ count = \displaystyle\frac{Associated\ device\ capacity\ (5/6\ GHz)}{Max\ associations\ per\ radio}$

In this example, you can simply calculate how many radios of a certain type will be required. Max associations per radio as stated earlier calculated at 60 users per radio for underseat placement of the access point or if mounted overhead, one radio per 200 clients.

Number of clients per access point

Clients per access point are both a design decision as well as a hardware limitation. While the following values represent maximum client association counts for certain platforms, designs should target significantly lower client counts to ensure optimal performance and user experience. You must always check the product data sheet to verify what the maximum allowed associations is per radio.

• AP-67x - max 512 associated client devices per radio (1024 total)

• AP-654 - max 1024 associated client devices per radio (2048 total)

• AP-518 - max 512 associated client devices per radio (1024 total)

• AP-574 - max 512 associated client devices per radio (1024 total)

• AP-635 - max 512 associated client devices per radio (1024 total)

The online LPV rough order of magnitude calculator by default calculates 60 client devices per radio when placed underseat and 200 client devices per radio when mounted overhead. These numbers can be tailored to the venue’s density requirements.

Bandwidth per client

Calculating the bandwidth per client is simply a matter of taking the channel’s available bandwidth and dividing by the anticipated number of clients per radio/channel.

This is not a perfect calculation and doesn’t take into account any of the real world conditions such as interference, congestion, distance, channel width, etc and one can easily take approximately 25% off that total because of lost airtime due to the nature of Wi-Fi communications. HPE Aruba Networks uses the term “Goodput” to define the actual amount of useable bandwidth minus the overhead, protocol limitations, and other factors like distance that can reduce the actual usable throughput.

Setting realistic expectations about Wi-Fi speeds is important - focus on whether the solution will provide a useable experience for the venue’s needs rather than theoretical numbers.

Associated device capacity

The associated device capacity (ADC) count and associated device per radio count are more important factors than bandwidth per client in designing a high-performance, high-density WLAN which meets the customers demands today and throughout the expected service life of the network. Depending on the type of venue, associated device capacity will vary. For example, a university lecture hall may have close to a 100% take rate versus a concert hall may have less than a 25% take rate.

This formula may provide a rough AP count, but will not guarantee a high-performance, high-density wireless LAN. This AP count still is not considering key performance metrics such as per-user bandwidth requirements or radio cell size. These two metrics can increase the number of access points truly required to deliver the correct design for the customer’s requirements.

In fact, high-density WLAN in large public venues must be designed with growth in mind. A high-density WLAN should be designed knowing that adoption increases over time and that the Associated Device Capacity number impacts much more than just the number of radios needed for a given band, but also all the upstream network infrastructure that will be required such as address space, ARP cache size, forwarding/bridge table size, DHCP lease binding database size, firewall sessions, public IP addresses for NAT/PAT, captive portal sessions, system licenses, HA dimensioning, and so on.

Other criteria such as the maximum associated devices on a particular radio can vary based on the model access point chosen.

Computing the total system throughput for a Wi-Fi network involves multiple considerations, including the Wi-Fi technology in use, the spectrum available, the environment, and user behavior. There are two main throughput numbers to be concerned with, Per AP Throughput and the Aggregate throughput across all APs.

To calculate a Per AP throughput, simply multiply the estimated per-client throughput by the number of supported users per AP. This can vary by AP Model and can be verified by checking the device’s datasheet.

Total system throughput

The basic mathematical formula to compute the aggregate throughput or the total system throughput is as follows:

$\displaystyle{Total\ system\ throughput\ (TST) = Channels \times Average\ channel\ throughput \times Reuse\ factor}$

Where:

• Channels = Number of channels in use by the high-density network.

• Average Channel Throughput = Weighted average goodput that is achievable in one channel by the expected mix of devices for that specific facility.

• Reuse Factor = Number of RF spatial reuses possible. For all but the most exotic high-density networks, this is equal to 1, which means no RF spatial reuse.

• Assume a 5 GHz deployment with 9 channels, an average channel throughput of 50 Mbps, and no RF spatial reuse, the TST would be 450 Mbps: $9\ Channels \times 50\ Mbps \times 1\ Reuse = 450\ Mbps$.

Additional channel throughput considerations

• Channel throughput depends heavily on the number of stations that attempt to use the channel simultaneously.

• Collisions and MAC-layer inefficiencies reduce overall capacity as more and more devices contend (compete) for medium access.

• Single-client throughput represents peak goodput performance measured during a speed test on a clean channel without other users present.

• Multi-client throughput is the weighted average goodput that is achievable in one channel by the expected mix of devices in a particular high-density area.

• Channel throughput can be further reduced by many impairments including misbehaving client devices, CCI, ACI and non-Wi-Fi interference.

• Real-world environments are rarely so pristine.

Another example calculation

Assume each access point can effectively deliver 100 Mbps of real-world throughput and support 50 users efficiently.

If 1,000 APs are deployed strategically throughout the venue:

• Each AP can support 100 Mbps.

• Total throughput for all APs is $1,000\ APs \times 100\ Mbps/AP = 100,000\ Mbps\ or\ 100\ Gbps$.

Then for each AP supporting 50 users:

• User throughput equals $100\ Mbps \div 50\ users = 2\ Mbps/user$.