Q. Does having multiple users create more of a problem? For example, we have a computer lab with 20 wireless laptops. Should I view each laptop as getting a 1/20th share of the bandwidth, or is it more like time-sharing, in which each of them taking turns at the whole 6M bit/sec?
I just read the Wizards’ Oct. 13 column on the network bandwidth tax. Does having multiple users create more of a problem? For example, we have a computer lab with 20 wireless laptops. Should I view each laptop as getting a 1/20th share of the bandwidth, or is it more like time-sharing, in which each of them taking turns at the whole 6M bit/sec? – James, Indiana
The answer to your question is “both.” Over the long term, each laptop would get 1/20th of the available bandwidth, assuming that all the laptops are trying to send a large amount of information at the same time. However, in the short term, each laptop would get the entire bandwidth of the WLAN for the time it takes to send a single packet, while taking turns. But there is more to this than meets the eye.
Your question addresses one of the dark technical areas of 802.11, the carrier sense multiple access with collision avoidance access mechanism (CSMA/CA). CSMA/CA is a probabilistic method used to spread out the times when access attempts are made. This spreading is invoked with every access attempt by every 802.11 device.
802.11 requires that a sending device must listen to the air traffic before beginning a transmission in order to avoid corrupting a packet currently on the air. This is the CSMA part of CSMA/CA. If a packet is detected on the air, the sending device must choose a random backoff time and wait for that backoff to expire before trying to send its own packet. A sender must also choose a random backoff time if its packet is not acknowledged by the receiver (usually indicating corruption or collision). Each consecutive time a packet is forced to back off, the maximum backoff time gets doubled. This is exactly the same method used in wired Ethernet environments.
Thinking this through, you can see that when there are only a few senders active, very little time (and bandwidth) is used for backoff, since most packets will be successfully received and acknowledged on the first attempt. However, as the number of active senders increases, more packets will wind up colliding and generating a second or third backoff before successful delivery to the receiver. Each of these additional backoffs adds exponentially more time to the transmission, and consumes more bandwidth.
To summarize, all 20 laptops will be sharing the air equally in the sense that they each take turns sending packets when air traffic is low-to-moderate. In this scenario, your time-sharing analogy works great as each laptop would experience the full 6M bit/sec of bandwidth when it was its turn to send a packet. But as air traffic increases, packet collision and resulting backoff times can cause more of a delay. If all 20 laptops were actively sending large amounts of information simultaneously, the laptops would, in effect, be sharing the 6M bit/sec of bandwidth together.
