[Battlemesh] [OpenWrt-Devel] chance to gather info from a high density deployment

[David Lang]

On Tue, 10 Feb 2015, John Crispin wrote:

> sounds scary, you should consider also mailing the battlemesh folks,
> even if it is not a mesh, they still have a lot of XP on how to measure
> "wifi"

I've added them. There are also differences between artificial testing and 
real-world deployments.

> assuming people have mobile phones, tablets and laptops that is an
> average of 50+ clients per AP  ... *yikkes*

surprisingly we see an average of about one MAC address on the network per 
person over the course of the weekend. I expected more myself. I guess that this 
means that people are interested enough in the talks and vendors that they don't 
pull out their gadgets :-)

But when you have ~400 people in a 4000 sq ft room (76'x45') you really can't 
deploy very many APs without having them fighting each other on 2.4GHz, so while 
the average clients per AP is low, it gets very high in spots.

see the La Jolla ballroom at 
http://www3.hilton.com/en/hotels/california/hilton-los-angeles-airport-LAXAHHH/event/roomcharts.html

It's not trivial, but I've been pulling it off for a few years now. The hotel 
upgraded to a 100Mb link last year which helped a bit (we phased out our heavy 
QoS shaping that we had been doing in the past)

David Lang


> On 10/02/2015 22:19, David Lang wrote:
>> I run the wireless network at the SCaLE conference which is taking place
>> next week. I will be deploying ~50 WNDR3700v2 and WNDR3800 APs running
>> OpenWRT to support ~3k users in the hotel (details available in the
>> presentation/paper I gave at LISA '12
>> https://www.usenix.org/conference/lisa12/technical-sessions/presentation/lang_david_wireless
>> )
>>
>> Since such high density environments are fairly rare and hard to get
>> access to, I want to give the OpenWRT project a chance to gather
>> information during the show. I will be building the image for this year
>> in the next couple of days, so I can even patch in additional logging of
>> data if needed. I will be building the firmware images in the next few
>> days.
>>
>> I gather syslogs and normal mrtg stats. I also gather the counts of
>> associations to each ESSID each minute (I use social engineering to
>> steer people to 5GHz by calling that network 'scale' and the 2.4GHz
>> network 'scale-slow')
>>
>> It was suggested that I see if I can gather
>> /sys/kernel/debug/ieee80211/phy*/netdev:*/stations/*/rc_stats and xmit
>> stats frequently (every 10s or so).
>>
>> What else can/should I gather?
>>
>> David Lang
>> _______________________________________________
>> openwrt-devel mailing list
>> openwrt-devel at lists.openwrt.org
>> https://lists.openwrt.org/cgi-bin/mailman/listinfo/openwrt-devel
>>
>