WiFi case (just rambling)

[Karl Klammer]

Hi,

I'll be providing wireless internet for a smallish open air festival this summer.
Approx. 1500 attendees on an organic farm, just outside of Munich.
I'm doing it just for fun/learning this time ... maybe for profit at some larger festivals next year.
How often does one get to play with large-ish wireless networks, without having any real pressure/responsibility? ;-)

I'm posting this as it involves two tough things, a cf19 mk3 and a pelican 1560.

The cf19 will be used as Performance Enhancing Proxy (PEP) with Polipo+500gb ssd for caching
and ssh tunnel to a remote server for bandwith compression.
This should help to push something like 50 mbits through a tiny 16mbits landline.
I reckon this won't handle more than a few youtube streams without buffering like hell,
but should be plenty for 500+ smartphones browsing the web and doing email.

I'll be using three Ruckus R500 for the three main areas (farm house + big tent + cow barn) with 5ghz backhaul mesh.
Two Ubiquity UAPs will be used for additional low-density backstage coverage,
while two Ubiquity Edgerouters will hopefully take care of all fixed-lan switching, routing and (basic) firewalling needs.

Anyways, I've assembled the case this week and want to share these observations:
1) a cf19 fits tightly into a pelican 1560, so this would be a great case for storing 8-10 cf19s.
2) Plasti Dip is excellent stuff for hardening & hiding pelican foam cubes, so that they won't tear easily.
3) A crazy amount of ethernet/power cables (~120m/~400ft) can fit into a tiny(38x15x15cm) spot ;-)

BR,
Karl Klammer

[Rob]

Network guy slobbering over here.......

GORGEOUS!

[Karl Klammer]

Glad you like it, Rob.

The Ubiquity stuff performs reasonably well at about 20% of the marketing specs IF you don't turn on to many features.
UAP and UAP-LR specs say 100 users, but performance goes down exponentially after about 20 active users per AP. (no airtime fairness)
EdgeRouter X can do ~950mbit switching, but only about 200mbit routing.
Ubiquities performance is "fair enough" considering the low pricepoint and root access.
Known technology, known limitations, will just work, not really interesting...


I'm really looking forward to giving that Ruckus stuff a serious testride.
I'm currently hoping(=designing) for 40% real world performance peaks, e.g. up to 200 instead of 500 active users per AP.
I will temporarily disable one or two live-APs during the event, if that is what it takes to get my benchmarking data ;-)

Some other Ruckus models are used by mobile carriers like Telefonica and Vodafone for Wifi-Offloading in crowded areas, e.g. shopping malls.
London Olympics also relied on them.

Here is a 45min semi-technical video on Ruckus antenna design:
https://www.youtube.com/watch?v=kcIkgyRGFQE

Some more semi-technical marketing pics follow...

[Karl Klammer]

I've finalized my setup and just packed it up for field deployment.
I got rid of the 2 low-density Ubiquity APs, as their space is now claimed by solar stuff.
One of my 3 high-density Ruckus APs will be put into tupperware and hung high up some tree, acting as a solar-powered mesh node

The rural internet uplink is only 600kbit up and 6mbit down , thus a lot of my focus shifted to WAN optimization techniques:
- Transport Layer Routing with Bandwidth Compression
- Hierachical Fair Service Queueing with TCP Ack Prioritization
- Web- and DNS- Caching with Filters for Ads, Autoupdates and Powerusers

pep01-pf.conf.txt

OpenBSD Performance Enhancing Proxy

(7.59 KiB) Downloaded 722 times

Some of these techiques are just plain silly and/or overengineered,
e.g. consider ICMP Ping behaviour enforced by pf.conf line 32:

Code: Select all

C:\Users\karl>ping 1.2.3.4
        Pinging 1.2.3.4 with 32 bytes of data:
        Reply from 1.2.3.4: bytes=32 time=1ms TTL=254
        Reply from 1.2.3.4: bytes=32 time=1ms TTL=254
        Reply from 1.2.3.4: bytes=32 time=1ms TTL=254
        Ping statistics for 1.2.3.4:
             Packets: Sent = 3, Received = 3, Lost = 0 (0% loss),
        Approximate round trip times in milli-seconds:
             Minimum = 1ms, Maximum = 1ms, Average = 1ms

C:\Users\karl>tracert heise.de
        Tracing route to heise.de [193.99.144.80]
        over a maximum of 30 hops:
           1    56 ms    <1 ms    <1 ms  10.42.242.1
           2     2 ms    10 ms     1 ms  redirector.heise.de [193.99.144.80]
        Trace complete. 

[Karl Klammer]

The festival happened this weekend.

I've learned why there is no such thing as a free lunch:
People don't appreciate stuff when they don't have to pay for it.

The orga decided to pull the plug and rather use up all the bandwidth for themselves,
about 5 hours after the first visitors arrived.

Oh well ... sometimes you win, sometimes you learn.

[Shawn]

Ugh