Isoroku <google@debabel.net> hath wroth:
>Greetings. I would appreciate some advice.
Free advice. Buy low. Sell high.
>I am hoping to provide free wireless access to a low income retirement
>community consisting of 100 small houses located 50 - 150 feet apart.
Ugh. Not good. Let's pretend you have a typical WISP (wireless ISP)
system with 100 users. Rule of thumb is that such a system can
support:
100 light email and web surfing users
10 business customers
1 file sharing user
Yep, 1 user can saturate your system with their file sharing habits.
So, you need some form of bandwidth control or bandwidth management.
So, back to the typical system. Broadband loading is currently at
about 10:1 or you can oversubscribe your bandwidth consumption 10
times assuming light users. You can look at this in many ways:
1. If only 10 users are on at one time, you have enough bandwidth to
keep the phone from ringing.
2. If all your 100 users are on at the same time, they will get
1/10th the peak available bandwidth.
3. If each user is expecting DSL performance, you need:
10 * 1.5Mbits/sec = 15Mbits/sec
backhaul performance.
4. If everyone gets online at the same time, 10% of them will ring
your phone with complaints.
>The terrain is extremely flat with only small trees, and the houses
>are all single level 2x4 construction, which seems to allow pretty
>good radio reception. But, I am also hoping to avoid the need for any
>tall antenna towers.
Towers are for covering large areas. Panel and sector antennas are
more effective, smaller, and more aesthetically correct. Think of
this as an exercise in lighting. You have a group of 100 houses that
need to illuminated? Do you do it with one huge unshielded light bulb
with no reflector that will light up the sky, the ground, and the
entire neighborhood? Nope. You get some spotlights in some key
locations, and light up houses in small groups. Look into sector
panel antennas:
<http://www.superpass.com/2400-2483M.html>
>I have no experience at this sort of thing, but my early research has
>led me to a new company called "Meraki" that claims to make it easy to
>spread affordable broadband access. Their "plug-and-surf" concept
>appeals to me, but at $49.00, the Meraki does not appear to be the
>best (lowest cost) solution for my problem. It seems that their
>repeaters are relatively low power, probably resulting in a need for
>one Meraki unit per home, and possibly even requiring some outdoor
>repeaters. (Not only would this be expensive, it would result in a
>large number of radio "hops" to the most distance users.) Therefore,
>I am now hoping to find a higher powered solution that could achieve
>similar results with less radios.
Meraki is a mesh network, where each radio also acts as a repeater,
thus eliminating the central access point or multiple access points
with expensive wired backhauls. Mesh networks are what is being sold
for municipal networks due to their relative ease of deployment and
lower price. Meraki is a commercialized spinoff from MIT Roofnet:
<http://pdos.csail.mit.edu/roofnet/doku.php>
I strongly suggest you read the following:
<http://pdos.csail.mit.edu/roofnet/doku.php?id=interesting>
which will give you a clue as to how things work. It's a bit complex,
but I can explain if necessary. The key points are that the
probability of delivery for a packet is roughly 50% or less for a
typical setup and that performance through more than 3 hops is
miserable.
>And, I have recently discovered the EnGenious ECB-3220 which claims to
>support "Wireless Distribution System (WDS)" at 400mW of output
>power. Could this be the right tool for the job?
No. WDS requires registering the access point that are connected to
the system in every other access point. There is limited table space
(32?) in most of these access point. You have to plot out the
topology in advance. It's also bridging, not routing, so you have no
isolation between clients. WDS allows connections for any client to
any other client. You don't need that.
>If configured properly, could multiple ECB-3220 devices provide a
>"Meraki-like" mesh network across the entire community that would
>allow shared internet access from a single DSL connection? (If
>necessary, I am willing to consider two DSL connections located at
>opposite ends of the community. But, I do not expect any heavy
>bandwidth demand. Just email and light surfing.)
For 100 users, I would think you should consider far more than 1 or 2
DSL connections. Never mind the lack of download performance. You
also have severely limited upstream performance with the typical DSL
connection. What will you do when the customer want to do VoIP or
file sharing that require more upstream bandwidth?
>Do you think that an inexperienced person like me (with reasonable
>effort) could learn to configure multiple EnGenius routers to work
>properly in "Wireless Distribution System" mode? Would two units be
>enough to test both my network configuration skills and the "WDS"
>capability of these devices?
Sure. It won't work, but you can do it. The problem is that such
systems were intended for comparatively small office LAN systems. They
don't scale well to 100 users. Neither do mesh networks or any other
system that relys on store and forward repeating. The break point
between a small "hot spot" type of system and a WISP system is about
25 users. To handle 100 homes (probably 200 computers), you should be
looking at Cisco, Sonicwall, 3com, or similar vendors.
>At $119.00 each on Newegg.com, it seems that the ECB-3220 would be a
>relatively low cost solution, and I am tempted to buy a couple or more
>to experiment with. What would be a reasonable range to hope to
>achieve between units? (500 feet?)
500ft will require directional antennas and line of sight at both
ends. The antennas are not horribly expensive, $40 for about 8dBi,
but will cause another problem. You can't easily build a mesh using
directional antennas. You'll probably notice that most of the
diagrams of mesh networks show omni directional antennas. That's
because you need to connect to anything in a 360 degree circle or you
get into weird topology problems or extraneous hops.
Also, it's easy enough to get WDS or a mesh network to function with a
small number of units. It's the inability for mesh networks to scale
well into large systems that cause problems.
<http://wifinetnews.com/archives/003972.html>
>I assume that testing may reveal the need to mount some kind of an
>external omni-directional antenna on the roofpeak of each house acting
>as a repeater. Any suggestions as to the best antenna?
Yes, but it's too soon for hardware selection. Work on the system
design and calculations first.
>By the way, I was thinking in terms of the EnGenius EUB-362 as the
>best wireless adapters for each end user in the community.
>
>Any and all thoughts, comments and advice will be greatly appeciated.
Yep. Start over.
1. Look into systems that are designed to provide wireless to a large
number of users. Most of them are based on a custom routing protocol
which is most important and vital. For starters see:
<http://www.mikrotik.com>
If the vendor cannot supply you with contacts that have an existing
working system in place with live customers, run away.
2. Forget about mesh networks. You don't have a large enough
coverage *AREA* to justify the hassles the come with mesh networks.
3. Calculate your backhaul usage. It's going to be your biggest
continuing expense. For 100 users, think multiple T1's, DS3, OC3, not
bonded DSL. Symmetrical, not asymmetrical. Also think in terms of
whether you need to buy and deliver routable IP's for users that want
to do VoIP and other services requiring route able IP's.
4. You need some way to manage this mess. You can't just deploy and
let it free run. You will have problems and it will be difficult to
find the cause (or culprits) without monitoring and control. Same
with adding and dropping user accounts.
5. Look into alternatives (HomePlug, HomePNA, running CAT5, data over
CATV coax, fiber, etc) as they are usually cheaper and easier to deal
with. If you have conduit available, definitely look into running
copper or fiber. If not possible, think about a tree distribution
topology, where you use wireless to link to groups of houses, and then
run wires from there.
6. You may need to hire a management service to run this mess. Do
you really want to get all the midnight phone calls from users? I do
this for a retirement home (60 residents). You would not believe the
strange calls I've received.
Some reading:
<http://www.bbwexchange.com/turnkey/>
<http://www.isp-planet.com/fixed_wireless/business/wireless_howto_p1.html>
<http://www.isp-planet.com/fixed_wireless/business/wireless_howto_p2.html>
<http://www.ispwireless.com> (mailing list)
--
Jeff Liebermann
jeffl@cruzio.com
150 Felker St #D
http://www.LearnByDestroying.com
Santa Cruz CA 95060
http://802.11junk.com
Skype: JeffLiebermann AE6KS 831-336-2558
A question for the experts out there... 
Linear Mode
