[CQ] Reality check on Amateur Mesh Networking
Bruce Perens
bruce at perens.com
Sun Oct 17 01:42:59 UTC 2010
I attended Jason's talk at Pacificon today. It seemed to me that there
was a concatenation of optimistic assumptions in the talk.
To get a reality check on mesh networking, please take a look at David
Rowe's "Mesh Potato" effort and his deployment of a mesh telephony
network in Dili, Timor Leste. In the post at
http://www.rowetel.com/blog/?p=652
Rowe discusses why a commercial Wimax device works well, even in Timor
Leste, when his own mesh network doesn't.
Also note that Rowe has created a 100% open mesh wifi design and is
having it manufactured, and has a history of success in manufacturing
Open Hardware - his Blackfin Asterisk PBX is great.
Jason showed a 150 mile wifi link as an example. While these things are
possible when all of the conditions are met, it's more the case that the
average ham won't achieve line-of-sight to even one other node. This is
the main reason that hams are still using D-STAR, AX-25, etc. We can do
better than D-STAR, potentially, with spread spectrum on lower
frequencies than those proposed.
Jason discussed mobile use of the mesh hardware. Has there been any
testing? I suspect that the multipath and fading issues might be
underestimated.
Jason offered TDMA as a cure for the hidden-node problem. While a good
algorithm for doing TDMA in a mesh is probably possible, it's not at all
clear that we have one yet. The RTS algorithm used on WiFi hasn't worked
well in practice on Rowe's mesh, in an area that didn't have much "Part
15" operation to interfere with it. Phil Karn KA9Q designed most of the
algorithm that was adopted in the WiFi standard, and I'm trying to get
him interested in why it's failing. I suspect that the well-working
examples have a star topology.
It's likely that we will lose 44 net in the coming IPV4 address panic.
Given the use we are making of it presently, it would be responsible of
us to give it up. Starting with IPV6 would make our lives easier over
the long run, there is an old TAPR paper on encoding Amateur callsigns
in IPV6 headers. If your local equipment only does IPV4, run NAT or map
the Amateur IPV6 subnet to a bank of net 10 addresses.
Codec2 was mentioned as a potential audio payload. Please note that the
7 byte Codec2 frame is tiny compared to the IP+TCP/UDP/RTP/STCP header
overhead, and that you either end up wasting most of the packet or
having really high latency because it takes 4 seconds of speech to fill
a packet. So, it turns out that a codec with less compression works
better over IP, Codec2 is intended to run with only one header per
transmission over the radio link.
I think there are some difficult, unsolved problems that should be
acknowledged in future talks. It might be best if the potential quoted
is what you achieve in your own testing.
Thanks
Bruce Perens K6BP
More information about the CQ
mailing list