[olug] Cox sucks

Wed Jul 18 03:40:57 UTC 2007

Sam,

I really appreciate your reply, as it has got me thinking.  To date, I
haven't done my due diligence crunching the financial aspects of such
an implementation, but your comments did get me Google'ing and I'm
beginning to see where some of your figures come from.  Some of the
things I've yet to fathom are the number of mesh nodes it would take,
and how many users each node could serve adequately.  However, like
cell-site deployment, I suppose a population density study would need
to be performed to determine the optimal placement for nodes, etc.

I guess my vision, initially, is to cater to the casual outdoor user
with a laptop/pda/etc sitting in a park, walking down the street, or
sitting outside on a balcony or backyard patio and that this user
desires to do some light web browsing, check e-mail, etc.  I'm not
looking to serve people pipes to facilitate first person shooter
gaming, VOIP, or video streaming.  By no means am I ambitious enough
to use 802.11x as a last-mile technology to serve home users.  If I
were to entertain the idea of a wireless ISP (WISP), I'd look into
WiMAX (802.16).  Of course, that technology isn't cheap and demands
CPE.

-----
"Topology issues would be next on the list as you mentioned mesh which
usually implies connecting multiple APs wirelessly for backhaul so you
further reduce your available bandwidth. Optimal would be a fully routed
network which would mean that data only flows through nodes that it has
to flow through, but this would negate roaming the mesh. If you allow
roaming then you can effectively halve the available bandwidth for each
node wirelessly linked together."
----

Yes and no.  The equipment I'm working with now are dual radio mesh
routers.  The radios are Ubiquiti XTreme Range "carrier class" radios.
 Radio 1 is an 11b radio acting as an Access Point, whilst radio 2 is
an 11a radio acting as a backhaul to other mesh routers.  Of course,
each router also has a wired ethernet port to route traffic over that
link in the event it is connected to a wired LAN to the Internet.  All
together, each mesh router costs around $700.  But if Nicholas
Negroponte can make laptops with mesh-capable wireless interfaces for
$100 (and I recall a company in India is aiming at making a laptop for
$50), I see no reason (even it it boils down to buying in bulk) that
this price couldn't come down substantially.

Each mesh router runs Linux and other open source (free) software to
reduce cost.  One interesting bit of software is OLSR (Optimized Link
State Routing) which runs on each router's 11a interface.  This allows
each mesh node to discover one another, setting up optimal routes to
the Internet.  Nodes without a direct (wired) connection use the 11a
radio to route traffic, while those with a direct connection use the
Ethernet interface.  The thing I like about this technique is that the
network topology is rather flexible, self-healing, and scalable.  If
you have an AP that is being over-loaded by users, or need to expand
coverage, stand up another mesh router in the vicinity of existing
ones.  Of course, a bit of graph theory is needed to ensure that node
failures don't isolate parts of the network.

I'll agree that, in this 2 radio scheme, you will see a 1/N bandwidth
fall off for each wireless hop that is traversed to get to the
Internet.  This is due to the inherent simplex nature of wireless
radio communication.  However, I'd propose a 3 radio solution which
would eliminate this phenomenon and provide full-duplex node-to-node
wireless links.  The author/owner of www.meshdynamics.com has some
good test data backing up my claim.

During my initial 2 radio testing, I've found that 3 wireless hops
provides "adequate" light web browsing, etc.  But 4 or more hops
degrades the bandwidth to a crawl and you start to experience TCP
timeouts, etc.

To sum it all up, I whole-heartedly agree that this is a huge
challenge, but that is what makes it fun to try to solve it.  I guess
I'm angling this concept from a technical (is it possible) perspective
and not the business (is it financially feasible) perspective.  There
are a host of issues (provisioning and deployment, security, quality
of service, etc.) that my project won't endeavor upon, but I still
feel that free outdoor wireless access to the Internet is a novel
concept worth exploring.

Ben

On 7/17/07, Sam Tetherow <tetherow at shwisp.net> wrote:
> Benjamin Watson wrote:
> > <SNIP>
> >
> >
> > On a final note, part of my thesis project is dealing with wireless
> > mesh networks.  Some articles/blogs 'round the 'net give a lot of
> > readers a bad taste when it comes to city-wide wireless Internet
> > access.  However, I think it is a novel concept which is gaining more
> > popularity.  I think Salt Lake City and a few other markets have
> > sponsored free wireless Internet access throughout their cities with
> > decent results and user satisfaction (I mean, its free, right).  My
> > research and testing shows that it is possible to stand up a wireless
> > Internet network blanketing a fair part of Omaha, but the problem is
> > access to an Internet backbone.  Grass routes efforts in other cities
> > are possible, where "regular people" stand up their own nodes and
> > share their bandwidth.  But in COX-town, I'd bet that wouldn't jive
> > with their ToS.  I wonder if Level3 wouldn't mind lighting up some of
> > the dark fiber they've got laying around all over for such an effort.
> >
> > Ben
> >
> >
> This is going to be a bit long winded....
>
> I have seen very few articles that report success for city wide wireless
> internet access, unless they are press releases from whoever is
> intimately involved with those networks. Both Earthlink and MetroFi
> announced that they are reviewing their business models and Earthlink
> has put on hold all new buildouts at this time.
>
> There are several problems with municipal level wireless, by this I mean
> city wide not municipal funded.
>
> Cost is one major factor, the current rule of thumb is between $200,000
> and $300,000 per square mile for capex and opex is estimated at 30%
> capex annually. This number is for infrastructure only and using these
> estimates you are still only reaching approximately 90% of all homes, of
> which only 10-15% can reliably connect without either an external or
> high powered CPE which will cost between $100-200.
>
> The next issue in a muni network is going to be available spectrum. It
> is going to be sharing 2.4GHz with every cordless phone, baby monitor,
> wireless entertainment center, security camera, microwave oven and home
> wifi router. If you move outside of 2.4GHz you will increase your
> infrastructure cost as well as requiring a CPE in 100% of all installs
> since laptops come with 2.4GHz standard all other frequencies require a
> seperate card.
>
> Next is the issue of bandwidth. Outdoor 802.11g has enjoyed very little
> success due to timing, power and sensitivity issues. In a best case
> scenario you are looking at 19Mbps optimal 802.11g throughput (at 54Mbps
> air rate) assuming a clean connection and 4Mbps optimal 802.11b
> throughput (at 11Mbps air rate), but given the nature of the network.
> You will most likely see suboptimal connections due to people who can
> only connect at a lower data rate either because the fall just above the
> 10% who can't be served or they fall in the 75-80% who really should
> have an external CPE but can connect marginally without it. Each user
> operating at suboptimal speed will effect every other user on that AP.
>
> If you abandon 802.11x for a proprietary protocol to address poor
> spectral efficiency you will increase the cost as noted above by
> requiring users to purchase external devices to connect to the network
> and you will still be limited by the theoretical 6bps/Hz maximum. In
> 2.4GHz that gets you 60*6=360Mbps theoretical maximum air rate across
> the entire band which will give you about 150Mbps total optimum
> throughput. You still have to deal with the effects of interference and
> poor connection quality though, which will significantly reduce this
> number. You can get access to another 60MHz of spectrum in the 5.8 ISM
> band but your propagation characteristics reduce further. There is also
> the possibility of even lower powered use of the 5GHz UNII band which is
> 265MHz but it is saddled with the new DFS requirement and as far as I
> know only one manufacturer has managed to get equipment through FCC testing.
>
> Topology issues would be next on the list as you mentioned mesh which
> usually implies connecting multiple APs wirelessly for backhaul so you
> further reduce your available bandwidth. Optimal would be a fully routed
> network which would mean that data only flows through nodes that it has
> to flow through, but this would negate roaming the mesh. If you allow
> roaming then you can effectively halve the available bandwidth for each
> node wirelessly linked together.
>
> I apologize for the length of this message but there are a lot of issues
> surrounding municipal level wireless networks and why they are such a
> challenge to implement.
>
> The only practical use I could see for a municipal level wireless
> network would be for the casual user so they can read email and lightly
> browse the web, but at $200-300K/sq mi upfront and $40-60K/sq mi annual
> recurring that is pretty costly. Once you get past the casual user, the
> network will not scale to handle the ever increasing bandwidth usage of
> the home user.
>
> Sam Tetherow
> Sandhills Wireless
>
>
>
>
> _______________________________________________
> OLUG mailing list
> OLUG at olug.org
> http://lists.olug.org/mailman/listinfo/olug
>