WiFi Airtime fairness

Fair AccessWi-Fi uses a shared medium and is used by many different client types and applications. So, fair access to send data becomes critically important, especially during times of high contention amongst multiple clients. First it is important that all clients receive access to that medium when they have data to transmit. At the same time, it is also important that newer, higher-speed clients can take advantage of that speed without being unduly slowed by older clients, which unfairly consume airtime. Further, some transmissions are intolerant to delay and must be prioritized ahead of data or background traffic.

Classifying Traffic
Most System uses a multistage classification system to determine traffic class and prioritization. Initial queue selection is performed by the  Policy Enforcement Firewall, and it can be determined in multiple ways, including wired-side QoS flags, WMM marks, or matching firewall policy. Traffic is then shaped on a per-virtual
AP basis, which allows fine-grain controls over traffic flows.
After traffic is classified and shaped, it enters a token-based queue. Based on priority and token availability, the frames are passed down to the WMM queues for transmission. The WMM system has two identical but separate sets of queues: one for high throughput (HT) traffic, and one for non-HT traffic. Figure diagrams the queue sequence.
Note: Voice traffic bypasses the fairness algorithm, because the time-sensitive nature of a voice call means that it must be placed directly into the queue without shaping.
Feedback comes from the radio unit to the token queue to influence the amount of traffic that is available to a client. Voice and video traffic and TCP acknowledgements also receive special priority. For voice traffic, the
system essentially cuts through the queues and provides strict priority. For TCP acknowledgements, the PEF module modifies the TCP window size dynamically to slow the sender where needed to avoid filling the queues.
Note: The fairness algorithm allocates tokens only to active clients. Associated clients that are not transmitting data will not have tokens allocated to them.

Fairness Options
The token queue is where fairness is enforced in the system. Most Vendor provides three options for deciding how
traffic fairness should operate:
 Default access: This option gives every queue equal weight, as would be the case for a WMM AP with no ARM algorithm. Default queuing may be preferred in some situations, but it results in less-capable clients getting more time on the air than faster clients.
Fair access: Tokens are allocated based on actual airtime used: clients that have used more airtime
recently receive lower priority for subsequent transmissions. The effect is to allow higher modes such as
802.11g vs. b to send more traffic in a given time interval. The result is “fair” in the sense that each client
gets equal time on the shared medium, independent of client type or capability.
 Preferred access: This option applies higher weights to faster modes. For example, this option assures
that an 802.11n client that can complete a transmission much faster than its 802.11a equivalent is given 
priority in the queue. Preferential fairness offers the highest overall data capacity, but at some cost to
less-capable clients. Some network managers use this option as a subtle nudge to the user population to upgrade to 802.11n clients. Some Vendors set time ratio 16:4:1 for n:g:b