Airtime Fairness

Airtime fairness is essential in wireless networks that must support critical enterprise applications. With airtime fairness, every client at a given quality-of-service level has equal access to the network's airtime. This is essential for ensuring predictable performance and quality-of-service, as well as allowing 802.11n and legacy clients to coexist on the same network. Without airtime fairness, organizations using mixed mode networks risk having legacy clients slow down the entire network or letting the fastest clients crowd out other users.

Though the concept sounds simple, airtime fairness has proven difficult to implement. Most networks give client devices control of their access to the radio spectrum. By nature, clients lack the network wide knowledge to allocate airtime fairly, and different driver algorithms implement different models of fairness that may not quite be so fair in the real world. Meru's Virtualized Wireless LAN technology is different. It gives the Meru network granular control over each client's use of the network. Other systems act like network hubs, so they are unable to allocate bandwidth to individual clients. With Virtualized Wireless LANs, an Ethernet-based design gives every client device gets its own dedicated Virtual Port.

Meru first introduced Airtime Fairness in 2003 and has been perfecting the technology since then, based on feedback from thousands of customers. Developed as a way to help smooth the upgrade from 802.11b to the then-new 802.11g system, it now offers three features critical to wireless performance:

  • Client Fairness

    Typical access points offer real TCP throughput of about 140 Mbps per radio in an all-802.11n network, with Meru's superior engineering achieving more than 190 Mbps. However, the performance figure that matters to clients is how much of that throughput is consistently available to each client. Airtime Fairness ensures that each client receives an equal proportion of the available capacity, with minimum throughput as close to the average as possible.

    For example, independent tests conducted by Novarum showed that when ten 802.11n clients were connected to a single Meru radio, each had an average TCP throughput of over 18 Mbps, for a total average speed of 180 Mbps. The minimum throughput achieved by any Meru client was 150 Mbps, showing consistent performance. This predictability makes it possible to guarantee that bandwidth will be available, which is critical if wireless is to replace wired.

  • Uplink vs. Downlink Fairness

    Most applications are either symmetric or require more downlink than uplink capacity; telephony and email send the same amount of data in each direction, while video streaming and web surfing involve more traffic sent from access points to clients than the other way round. This means that an access point needs to transmit far more data than each individual client does, with the ratio growing as more clients are connected to each access point. Hub-like access points are unable to accommodate this because they treat the access point's radio as just another client contending for network access.

    Meru is different. The Meru controller ensures that capacity is shared equally between upstream and downstream transmissions when both need to be sent, ensuring optimum performance. This is particularly important for voice traffic, where even a few tens of microseconds' delay in transmitting a packet can make a conversation unintelligible. Where other networks sacrifice downstream for upstream performance, Meru ensures that calls get through in both directions.

  • 802.11n vs. Legacy Fairness

    When all clients are able to transmit at the same data rate, Airtime Fairness gives each an equal share of the total capacity. However, wireless networks often involve clients that are capable of a wide range of data rates, due to both differing signal strength and the need to support legacy devices. The 802.11n standard supports a range of data rates from 300 Mbps to 1 Mbps, resulting in highly variable performance. If every client was allowed an equal share of the available capacity, the slowest clients would dominate the network because they take longer to transmit a given amount of data.

    With Airtime Fairness, each client experiences throughput proportional to its data rate. 802.11n clients see the same throughput that they would if connected to an 802.11n-only network, while legacy clients behave as if connected to a legacy network. As in switched Ethernet, each client is able to negotiate its own data rate with the network and is not impacted by others.