Server Fault is a question and answer site for system and network administrators. It's 100% free, no registration required.

Sign up
Here's how it works:
  1. Anybody can ask a question
  2. Anybody can answer
  3. The best answers are voted up and rise to the top

Background: I'm in a noisy environment, and I'm trying to optimize my WiFi network to have a more stable connection for the somewhat high volume of users (~50-75 on a busy day). There are 4 APs, and I've already adjusted the channels and transmit power, and overall I have fairly decent coverage. However, I still get about ~10% packet drop when pinging Google and walking around the building, roaming from AP to AP.

In most WiFi APs I've seen, the default RTS Threshold is set at 2347 (from what I've read in various places, this setting counts as "disabled"), and Fragmentation Threshold is set at 2346. My particular brand of router is set at 2346 and 2346. I have several questions...

  1. Where is the value of 2346 derived from? It seems somewhat arbitrary, however, the notes for Frag. Threshold indicates that it needs to be over 256 and an even number.

  2. How are the RTS and Frag. Thresholds related? Their values can't be coincidence.

  3. If modified, should they be changed together?

  4. What's a safe value to try lowering them to, for starters?

My priority is not necessarily getting peak bandwidth for each device, but giving users a stable, consistent bandwidth/connection.

share|improve this question
Are you running a mixed b/g network? If so, that may account for a lot of the issues. – Greg Askew Mar 15 '12 at 1:13
Yup, and there's no way to disable B or set a minimum data rate on these devices. – Bigbio2002 Mar 16 '12 at 18:07
up vote 10 down vote accepted
  1. 2346 is the maximum 802.11 frame size. Setting the RTS and fragmentation thresholds to the maximum means that no packets will meet the threshold.

  2. The fragmentation threshold limits the maximum frame size. This reduces the time required to transmit the frame, and therefore reduces the probability that it will be corrupted (at the cost of more data overhead). The RTS threshold specifies the frame size at which the transmitter must use the RTS/CTS protocol, which is largely to solve the hidden node problem. This obviously also adds overhead.

  3. Not necessarily -- if you don't have a hidden node problem then changing the RTS threshold won't improve performance. In order for RTS/CTS to kick in the RTS threshold must be the same as or smaller than the fragmentation threshold though.

  4. I would start with setting them such that a standard Ethernet frame is fragmented into two 802.11 frames (1500/2 = 750 bytes payload + 34 bytes overhead = 784 bytes) and any frames bigger than a third of a standard Ethernet frame uses RTS (534 bytes).

As far as I know though, both these settings only affect the transmitter, i.e. configuring them on the AP only makes the AP use them for its transmissions and doesn't make the clients use them for their transmissions.

share|improve this answer

That mixed b/g scenario is particularly suboptimal. You may want to review some of the previous discussions on the topic, such as:

Slowest wireless client dictates the connection quality of all others?

Also, another performance killer occurs when point A can receive point B's signal, but B cannot receive A's signal. Someone else on ServerFault pointed this out as the "hidden transmitter effect". More about that phenomena at the link below. They point out that:

"...While horizontal polarization is desired, the lack of inexpensive commercial horizontally polarized omni-directional antennas may require the use of vertically polarized antennas. A good omni- directional vertically polarized antenna will cost about the same as a parabolic antenna. Use of an omni-directional antenna helps minimize the "hidden transmitter" effect. "

share|improve this answer

Your Answer


By posting your answer, you agree to the privacy policy and terms of service.

Not the answer you're looking for? Browse other questions tagged or ask your own question.