Hello!

There was a very interesting question. Against the background of the fact that in all the Wi-Fi documentation, not a word was said about the transfer of data from the end point to the end point, we decided to test it ourselves. It did not bring us any special results, so I was carried to you.

So, everyone knows that (suppose the channel is free) before transmitting data, the station waits for DIFS time, listening to the channel, if there is nothing in the channel during this period, it starts counting the random time t (backoff). Only after that we have the opportunity to send data. In addition, in order to take precedence over the others, there is a SIFS wait time, which is maintained before sending an ACK message from the access point to the station.

This is followed by a part that has not been described experimentally obtained:

But, before transmitting the frame received to the buffer of the access point to the recipient, the access point again performs DIFS + t (backoff), which aligns it with regular users and does not give any advantage. Accordingly, if each time t (backoff) of the access point is greater than any of the transmitting stations, it will slowly fill its buffer, but will not give any data at all. That, in our opinion, is not a very good solution, because memory is finite, and in the worst case, the network can stop transmitting data altogether. In terms of logic, would it not be better to use SIFS before sending a frame to the end station?

Actually the question is, are we right in the reasoning and tests? If not, please confirm with literature.

    2 answers 2

    The Infrastructure Mode (with AP) uses the Point Coordination Function (PCF). See IEEE 802.11-2007, Section 9.1.2 [1].

    1. http://standards.ieee.org/getieee802/download/802.11-2007.pdf
    • It can be used, but in fact is not used, not to mention> This access method uses a PCT. What is the conclusion, using only the access point and N clients, it will act completely equivalent to them, and therefore using the bug in the firmware (the choice of the smallest t (backoff)) you can bring down the network? In addition, there is a second case when one client gets to the Internet through TD. Having t (backoff) less than the point, he can send packets, and the point can not return it to him. - Dex

    Maybe I'm wrong, but it seems to me that any network device has some time to fill the buffer. And if the buffer was not filled for a certain period of time, then what was accumulated earlier is transmitted. That is, the data in the buffer can not hang forever, waiting for the buffer to be completely filled.

    • Do not quite understand what you mean. Apparently, you did not understand the question. This refers to the frame buffer (frames) and the "artificial" filling of the access point buffer in a situation where its t (backoff) is always greater than the stations connected to it. - Dex