How many wireless devices can connect to an 802.11g access point?
The number of clients that can simultaneously connect to a single 802.11 AP depends upon a multitude of factors.
To start, the AP may have a hard-coded or licensed limit on the number of client associations. For example, the Linksys WCG200 connects up to 32 clients, while the Ruckus ZoneFlex 2925 connects up to 50 clients.
The AP may also have limits based on the type of security that active clients are using. For example, the ZoneFlex 2925 supports up to 50 open, WEP, or WPA-AES clients, but just 22 WPA-TKIP clients due to the inherent inefficiencies of the Temporal Key Integrity Protocol (TKIP ).
Every AP is limited by number and type of radios and channels used. Channels are shared resources, so the sum of all traffic sent and received on that frequency cannot exceed that channel's capacity. Aggregate application throughput is further diminished by 802.11 protocol overhead, client competition, and loss/retransmission due to distance and interference. To illustrate this point: an 802.11g D-Link DI-524 that delivered 20 Mbps to a client 6 feet away dropped to 10 Mbps at 110 feet (see this PDF to view the OctoScope test). But an 802.11n D-Link DIR-655 delivered almost 130 Mbps at 6 feet and 40 Mbps at 180 feet when measured in the same environment, using the same applications. Furthermore, different client cards often have different experiences.
Throughput, latency, and application needs directly impact the number of wireless devices that can effectively use a given AP. Transactional data applications are less bandwidth-intensive than video applications, and both are more forgiving of delay than voice applications. For example, the 50 data clients supported by the ZoneFlex 2925 drops to 20 concurrent voice over Wi-Fi calls or dozens of IP video streams, depending on stream encoding and image resolution.
The numbers given here are just a few examples, drawn from spec sheets and competitive performance tests. Bottom line: the "max clients" line item on any AP's spec sheet is an upper-bound. Actual limits must be measured in your own environment, with a representative application and client mix.
This was first published in July 2007