From mboxrd@z Thu Jan 1 00:00:00 1970 From: Ali Abedi Date: Wed, 31 Oct 2012 14:54:44 -0400 Subject: [ath9k-devel] Channel busy cycles In-Reply-To: <509168B9.7000508@mailservices.uwaterloo.ca> References: <509168B9.7000508@mailservices.uwaterloo.ca> Message-ID: <509173F4.7070502@mailservices.uwaterloo.ca> List-Id: MIME-Version: 1.0 Content-Type: text/plain; charset="us-ascii" Content-Transfer-Encoding: 7bit To: ath9k-devel@lists.ath9k.org Here is an update from me. I added a round-robin rate adaptation that cycles through all 11g rates and I measured busy - (TX-RX) for all rates. This measure (busy - (TX + RX)) does not decrease as we increase transmission rate all the time. There are anomalies like rate 18. Please see the plot. config: 802.11g, 1 sender 1 receiver, channel cycle information is collection at sender. Retry 1. RTS/CTS off. I can generate the plot for RTS/CTS on, let me know if you want to see it. Thanks, Ali On 10/30/2012 09:29 PM, Adrian Chadd wrote: > I've asked; SIFS is the only thing that's come back thus far. > > > > Adrian > > > On 30 October 2012 18:24, Ali Abedi wrote: >> Hi Adrian, >> >> Thanks for asking. Is there any update? >> I think NAV is not the issue at least in our current setup, since we have >> only one sender. And I collect channel cycle information at the sender, so >> the only thing it receives are acks and beacons. >> >> Thanks, >> Ali >> >> >> >> >> On 10/28/2012 11:54 PM, Adrian Chadd wrote: >>> On 28 October 2012 20:49, Wright, Brett >>> wrote: >>>>> On 28 October 2012 15:12, Ali Abedi wrote: >>>>>> Thanks for mentioning this. It is important to check these things, >>>> but >>>>>> it did not affect the numbers. I really suspect that some portion of >>>>>> frames are considered "busy" but not tx or rx. >>>> I was thinking that perhaps busy includes virtual carrier sense (i.e >>>> NAV) as well as busy due to CCA/noise? >>> It does include CCA/noise. No idea about VCS/NAV. I thought it was >>> controlled by a line from the BB (RX_CLEAR) rather than VCS/NAV, which >>> is a MAC counter function. But I could be wrong. >>> >>> I've punted the question internally; I'll post back when I have more >>> details. >>> >>> These can be easily tested though - just fake some NAV entries and see >>> if your busy counters go up. >>> >>> >>> >>> Adrian >> -------------- next part -------------- A non-text attachment was scrubbed... Name: plot.pdf Type: application/pdf Size: 160920 bytes Desc: not available Url : http://lists.ath9k.org/pipermail/ath9k-devel/attachments/20121031/e935b84c/attachment-0001.pdf