[Cu-wireless] HSLS?

[Todd Boyle]

Hi Sascha,

Given the following circumstances what would be the
performance of the CWN (latency in ms, jitter, and
thruput in kbps) ?

User geographical locations:

draw a checkerboard grid of 100 ft. squares.
put a house in the middle of each square.

choose a random selection of 10% of the houses.
put the standard CWN node/omni on this 10%
Pencil in links between all of these CWN participants
if they happen to be within 250 feet.
-
That's the installation pattern if we have a wildly
successful, 10% penetration as a new ISP provider.
---

Now, erase 50% of the links other than contiguous
properties, because of trees or other obstacles.

choose another random selection of 20% of
the houses.  Put in a typical noisy 2.4Ghz AP
or camera or other device as intermittent,
interference sources.

---
My question is, what will be the latency, jitter and thruput
when 50% of the users are online in the evening, at a moment
when 50% of those users click DOWNLOAD for something?

(you notice, my focus on broadband speed.  I belive there
is great value in a P2P mesh that goes as many as 20 hops,
or indefinitely like FIDONET, just for email! but most people
will not buy it.  Their expectations are higher than that.)

--
The answer of course depends on where the Internet
gateways are located.   So the question becomes a
circular question. What we're really asking is how many
gateways do we need to install?  One at ever alternate
node?  In that case we could just use one AP per gateway.
--
(So the question becomes is it even worth it to have
a mesh that transmits ONE extra hop? i.e. a repeater?
Of course it's worth it!  I'm sure it is.. )

----------

I know how much the devices will cost.  Something
like $400, plus, they will not be five nines, they will
have maintenance events that have economic cost
certainly, greater than $100 to resolve.

But I don't know what the performance picture to
tell the users, when I try to convince them to layout
$400 for these devices.  I have been told by numerous
people that the performance with a sinlge omni
would suck, in these circumstances without bigtime
(proprietary) engineering at the physical layer
(the firmware in the radios) to schedule transmissions,
dynamically change routing costs at milliseconds
intervals, negotiate contention, downshift the power
levels opportunistically, all that stuff.

Todd Boyle - Kirkland WA - 425-827-3107
http://www.ledgerism.net/ , http://refusenik.org

-------------- next part --------------
An HTML attachment was scrubbed...
URL: http://lists.cu.groogroo.com/mailman/archive/cu-wireless/attachments/20040309/e0e710ee/attachment.html