Hi Dave,
Nothing personal but I predict that Mesh Networks will find (after several
deployments)
that the actual throughputs through 3 mesh nodes will be significantly less
than the throughput over a well designed single hop (pt-multipoint)
network. From beginning node to end node through 1 intermediate mesh node (2
time slots for
the intermediate node) the throughput will be down to about 45% of what it
would be over a direct (single) link.
With 2 intermediate mesh nodes, the throughput will be down to (.45 x .45 =
20%) 20% vs. a single direct link.
With 3 intermediate nodes, it's down to (.45 x .45 x .45 = .09) 9%.
(I've used .45 based on my experience deploying PC-based wireless routers with
a single wireless port.
The actual throughput was slightly less than half of the throughput without
repeating.
Further, when we compare throughput to raw data rate; the actual throughput is
about 50-55% of the
raw data rate. An 11 Mbps "raw" link will have an actual throughput of about
5.5 - 6.0 Mbps.
If we now multiply 6 Mbps (using the more optimistic end of the range) by 9%,
we get an
actual throughput of 540 kbps through 3 intermediate mesh nodes.
Finally, even this figure is too high. In the real world, there will be less
than perfect signal isolation
between nodes that are separated by only one intermediate node. This means
(remember,
all the nodes are on the same frequency) that there will be some packet
collisions between nodes
with resulting retransmissions. These collisions result in wasted timeslots
which causes even lower throughput.
For example, node one packets will likely collide occasionally with node three
packets and
node four packets may collide with node 2 packets. The resulting
retransmissions will substantially
reduce the throughput below 540 kbps.
In conclusion, IMHO, it is very unlikely that any single-frequency
(single-wireless-port nodes)
mesh scheme will ever serve to deliver serious bandwidth.
It's a shame really because the node-to-node-to-node concept is so attractive
on an
intuitive level (teamwork, bucket brigades, person-to-person handoffs, etc).
On the
single-frequency RF level, alas, it may never work. Now, a two or
three-wireless port
node is a completely different story.
Best regards,
jack
Dave Close wrote:
> Frank Keeney wrote:
> >As you know MeshNetworks is not compatible with 802.11b
>
> Mesh Networks exhibited at the 802.11 Planet conference back in December
> and I spoke with their people at length. They use 802.11b so they are
> certainly compatible with it. It is true that cards without their features
> would not benefit from the mesh, but they would work.
>
> At present, they are targeting "campus" or "building" networks, not open
> metropolitan networks. Much of their benefit comes from keeping all the
> radios within a close distance so that 802.11 operates at its peak speed,
> instead of degrading to a lower speed to accommodate a weak signal. Even
> relaying a message three times at 11Mbps is faster than passing it once
> at 1Mbps or 2Mbps.
> --
> Dave Close, Compata, Costa Mesa CA +1 714 434 7359
> [EMAIL PROTECTED] [EMAIL PROTECTED]
> "Quantum computing is a marvelous way to show the non-
> intuitive nature of quantum mechanics." -Gordon Moore
--
Jack Unger, President-Wireless InfoNet ([EMAIL PROTECTED]) 818 227-4220
Vendor-Neutral WISP Training - http://www.ask-wi.com/2002workshops.html
Author - "Deploying License-Free Wireless WANs"
http://www.amazon.com/exec/obidos/ASIN/1587050692/
