Ok I didn't realize that you where utilizing the WDS station mode. So
you basically have a normal AP/station setup but it is just bridged.
Are you using N-stream? The WDS-station mode really was designed so
that N-stream could be used on a WDS / Bridge network. WDS - station is
a proprietary mode developed by MT if my understanding is correct. If
so the likely the issue will be resolved with the polling feature
available with N-stream. The reason I state this is because from the
information provided the issue has became a problem as more load has
been applied to the solution. The solution is more then capable of
handling the throughput so this would indicate an interference source.
As 802.11 is the solution you are seeing more retransmissions as the
wait-before-talk mechanism is causing high latency issues. Assuming the
interference is self generated and antenna choices are limited the
Polling feature in N-stream likely is your best bet for fixing this.
I currently use N-stream over WDS for one of my main back hauls to a new
bandwidth source and it has performed flawlessly for 6 months. This is
using 2.9.28 software. It has been upgraded since installation and I am
not sure what version of the software we started with.
Anthony Will
Tom DeReggi wrote:
To be clear, Mikrotik us being used, and the 4 remote building are in
wds station mode and only configured to talk to the 1 central master
WDS AP, the four client WDS radios are not configured to talk to each
other. So all the CPE radios only have one hop to the APconnected to
the Internet backhaul.
My theory for design was...
I had a 10 mbps backhaul. The WDS PtMP would have 16mbps (54 mbps
modulation), to help with waste from re-transmissions. All clients are
bandwidth managed (priority weighted method) centrally on other end of
backhaul, to also assist with fair transmission time. Also radios use
CDMA/CA, with the CA also assisting. The question is, is this enough
to let it work well with only four buildings.
I'm starting to think that it might not be. But the problem shouldn't
be that they hear each other. we want them to hear each other, so they
don't transmit at the same time. Thats what 802.11 needs. Hidden node
happens because CPEs don't hear each other, and don;t know someone
else is transmitting, from my understanding.
Part of my question is, Does WDS work differently when in Mikrotik
Station WDS mode than a normal WDS AP?
Tom DeReggi
RapidDSL & Wireless, Inc
IntAirNet- Fixed Wireless Broadband
----- Original Message ----- From: "Anthony Will"
<[EMAIL PROTECTED]>
To: "WISPA General List" <[email protected]>
Sent: Saturday, October 07, 2006 11:57 AM
Subject: Re: [WISPA] WDS PtMP
It would seem to me that as your load increased your WDS/APs are
transmitting over each other as clients are trying to transmit to the
central AP. client -->WDS/AP transmitting carrier beacons or other
data to client and passing onto to -->WDS/AP<--WDS/AP<--Client
(transmitting to local AP)
In this scenario you have the two clients talking and one AP all
trying to talk at the same time and thus raising your noise floor
because they are all on the same channel.
There is not a feature in standard WDS to coordinate who can talk and
who can not talk other then the standard CDMA layer of the 802.11
protocol. This will create issues as the more load you have on this
setup the more self interference and retransmissions you will incur.
The big thing the mesh brings to the table is the ability to help
coordinate all of this traffic so that you can utilize the spectrum
more efficiently. At least that is my opinion as soon as someone
actually does it. You likely are going to have to switch to a
station /AP solution for this setup because everything is to close
and can hear each other. This will destroy your bridge setup unless
you change to a propitiatory system such as Trango, Canopy, etc. One
other thing to note is that this is all half duplex so you might have
two many hops and thus running out of bandwidth.
Anthony Will
Broadband Corp.
Tom DeReggi wrote:
Background....
In standard WIFI, a principle exists called hidden note, where two
CPEs transmit at the same time and colide because they do not hear
each other. There are three ways to get around that, using WIFI
between Client and AP. 1) Polling (Karlnet, Nstream, Proprietary),
2) Use Omnis, so radios can hear each other if in close proximity,
3) RTS/CTS which effectively solves the problem at a significant
performance degregation. A well know problem with well known
solutions.
Issue.....
How does this play our with WDS? AP to AP communication. Sure in PtP
its a non-issue, because there are only two radios involved to
complete the link. But WDS allows PtMP operation.
How does WDS commuication work? Does the Hidden Node problem exist
with PtMP WDS? And if so, is there a way to address it? If so, will
it help to make the CPE's Omnis, so they hear each other?
My confusion is how WDS/WDS works compared to Station/AP modes.
Example application:
Using 802.11a gear.
5 seperate MTU buildings, spread out within 300 yards of each other.
1 is a Master AP Site, with an Omni, and a second backhaul radio to
the Internet.
4 of the 5 have a direction CPE style antenna pointing to the Master
Antenna.
WDS is used to allow the radios to operate as true transparent
bridges, and to pass per client (5-10 clients per MTU) large packet
VLAN traffic.
(Note: There is a reason we did not select Nstreme w/ Polling. It
may have been an incompatibilty with WDS or inabilty to do
transparent bridging with large packets, which standard 802.11
station mode does not support under protocol. May have been early
version of Firmware, not sure if still an issue)
Why I thought it might be an issue:
Surveys show low noise. However, as more clients have been taken on
(2 mbps average sustained throughput all combined), the Link quality
started to degregate as if the noise floor was rising.
As a tempoirary measure, we switched to 5.2Ghz (indoor only FREQ,
which appeared not to have any detectable noise in standard 802.11
based survey tools, and was chosen because non-detectable carrier
grade gear would not use those channels). Its hard to believe that
the noise floor would be that high using that freq. So I'm
wondering if the noise that I'm hearing is actually my own CPEs
within this project?
The symptom was sparatic higher latency, what typically would happen
if 802.11a had frequent retransmissions (native prorocol ARQ). I can
look at stats to see if there are re-transmissions, but that data is
pointless, as what I want to know is, is the retransmisison because
my own noise or someone elses. Its hard to tell with WiFi, as WiFi
doesn't transmit when its not in use. So testing in the middle of
the night, when clients and users in town are off, may not be
meaningful. Its also possible, that I just have a failing radio
card or two, and a totally different cause.
Tom DeReggi
RapidDSL & Wireless, Inc
IntAirNet- Fixed Wireless Broadband
--
WISPA Wireless List: [email protected]
Subscribe/Unsubscribe:
http://lists.wispa.org/mailman/listinfo/wireless
Archives: http://lists.wispa.org/pipermail/wireless/
--
No virus found in this incoming message.
Checked by AVG Free Edition.
Version: 7.1.407 / Virus Database: 268.13.0/465 - Release Date:
10/6/2006
--
WISPA Wireless List: [email protected]
Subscribe/Unsubscribe:
http://lists.wispa.org/mailman/listinfo/wireless
Archives: http://lists.wispa.org/pipermail/wireless/
