Close

HPE Aruba Networking Blogs

Contention Delay Killed the WLAN Star

By Andrew vonNagy, Blog Contributor

The Wi-Fi industry seems dominated by discussions on the ever-increasing bandwidth capabilities and peak speeds brought with the latest product offerings based on 802.11ac. But while industry marketing touts Gigabit capable peak speeds, the underlying factors affecting WLAN performance have changed little.

802.11ac does bring modest gains in speed through higher order modulation with 256-QAM, but the practical limitations of its use greatly reduce its benefit. The bulk of recent improvements in peak speeds are not due to some magical advancement in RF capabilities that grant us new bandwidth or capacity, but stem from the more mundane explanation that we are simply using the spectral resources we have in a different, re-arranged, fashion through wider channel widths. With wider channel widths we are effectively "robbing from Peter to pay Paul." That is, we are stealing spectrum from neighboring APs in order to increase potential peak speeds any single AP.

This new arrangement can work well in consumer, home, and rural applications where AP density is low enough that wider channels allow us to utilize dormant spectrum that we weren't using prior. However, in enterprise environments where all available unlicensed spectrum is already being utilized we have no such luxury as dormant spectrum. Using wider channel widths may or may not make sense based on a number of factors, ultimately boiling down to the resulting effect on medium contention.

Medium contention is the true driver in the success or failure of a WLAN and we must effectively understand its effect on WLAN performance in order to design and optimize our networks.

A Framework for Network Performance

Let's begin by providing a frame of reference for network telecommunications performance in general. The two largest factors in network performance are bandwidth and latency (also referred to as delay). These two are inextricably tied together. We increase bandwidth through improvements that allow higher speeds (through lower packet serialization delay) and more data in transit at once (for example leveraging lower end-to-end delay with larger TCP window sizes). We reduce latency by minimizing geographic delay, serialization delay, and contention delay (as Martin Geddes has expertly explained). Stated another way, we increase bandwidth by reducing the amount of time it takes to send bits from one point to another.

GSV graphic.png

Network Sources of Loss and Delay, courtesy of Martin Geddes

We are at a point where we've squeezed most of geographic delay we can out of the system since communications are transmitted near the speed of light. We have also reduced serialization delay (and improved bandwidth speeds) by enormous amounts to the point at which there is little gain left to be realized. We can see an example of this with 802.11ac; the use of 256-QAM over a highly variable and lossy wireless link is extraordinary, but the practical use of this higher packet serialization rate (what Wi-Fi engineers refer to as modulation rate) is limited to a range of few meters from the access point. Truly little additional gains can be realized in this area.

Serialization Delay.png

Reducing Serialization Delay, courtesy of Martin Geddes

So what is left to improve network performance is the reduction of variable contention delay.

Factors Affecting WLAN Contention

In order to optimize WLAN performance we need a thorough understanding of the factors affecting medium contention. This requires Wi-Fi engineers to focus on evaluating airtime demand by clients, optimizing cell density (clients per radio), using all available spectrum, and carefully designing frequency re-use.

Factors of Wi-Fi contention include:

  1. Airtime Demand – the amount of time that each client or AP requires in order to transmit the data required by one or more applications. This is largely a function of AP and client capabilities, application throughput and packetization characteristics, and resulting spectral efficiency.
  2. Cell Density – the number of transmitters within a Wi-Fi contention domain (frequency or channel) and their airtime demand (probability of transmission) affects the frame error and retransmission rate. The goal is to optimize cell density to efficiently utilize AP and channel resources without overloading a Wi-Fi cell and causing significant contention induced performance degradation.
  3. Spectrum Inventory – the number and width of Wi-Fi channels available that will be used to segment users into different contention domains to avoid sharing airtime and capacity. This one is pretty simple to understand: the more spectrum we have, the more we can segment users, co-locate APs if necessary, and increase aggregate WLAN capacity. A more nuanced examination includes analysis of tradeoffs between the number of channels and channel width to optimize WLAN performance and capacity.
  4. Frequency Re-Use – the ability to effectively re-use Wi-Fi channels to avoid co-channel interference (CCI), which is another form of contention induced performance degradation.

We use capacity planning to model airtime demand, provide the appropriate quantity of access points that optimizes cell density (clients per radio) to prevent contention induced performance degradation, and perform what-if scenario analysis related to spectral efficiency (such as channel width permutations). This is the foundation of the Revolution Wi-Fi Capacity Planner tool.

We use RF planning to leverage our spectrum inventory into a design that provides optimal RF frequency re-use which allows APs to co-exist without causing co-channel interference (CCI), which is another form of contention induced performance degradation. This includes channel and transmit power planning, AP placement, and appropriate antenna selection to focus signal propagation.

Finally, we must integrate capacity and RF planning together in an iterative design approach to achieve a final WLAN design that provides sufficient coverage and capacity to meet the needs for each unique network environment.

In the next article, I'll dive deeper into the mechanics of client airtime demand in order to better understand WLAN capacity planning.

Cheers,

Andrew von Nagy