On Sun, 7 Aug 2005 19:03:10 +0000 (UTC), rel <relic@home.nl> wrote:

>let me first say that I know little about wifi and the situation makes
>me nosey at what is going on.

I'll try to answer your questions, but this is much too messy a
problem to solve with what information is supplied. Literally
everything you've supplied about the system makes me wonder what your
wireless experts were thinking.

(Slight redraw. Beware of tabs.)

> A1 B1 B2 C1
> ______|< >|_____|< >|_____
> | | | | | |
>
>Building A Building B Building C
>A1 = Redline B1 = Redl. C1 = Cisco 350 bridge
> B2 = Cisco 350 bridge
>
>So building B is higher than building C, that is why this 3 point
>setup is used.

So far no problems as long as you have line of sight. However, I'm
worried about interference pickup from the rooftop. The Redline
bridges are supplied with their own directional antennas, so I'm not
too worried about them. However, you didn't mention what the Cisco
350 bridges used for antennas. Hopefully they're directional and not
an omnidirectional.

>For what I know devices B1 and B2 are connected with
>a 75ohm COAX cable that spans the width of the the B building whitch is
>aprox 60 meters. My guess is that signal strenght is lost here.

There's no RF in the coax cable. The coax possibly carries only data.
I think you should check if it really is coax cable. There's no
evidence of any media converters in this system to go from 100baseTX
to 10base2 coax. In addition, the coax cable should be 50 ohms and
not 75 ohms. I've used 75 ohms, but only if there was no alternative.
Shielded CAT5 (required in some EU countries) looks very much like
RG-6/u 75ohm coax. Check the labels.

>> Yeah, I know RedLine claims that the AN-50 is NLOS because it does
>> OFDM. I think NLOS is science fiction and marketing hype.
>
>I've read on their site that they develop their systems by
>a mathimatically view on the concept. And so be better :)
>Aren't they all since it's all about EM waves.

In a past life, I played RF engineer for about 15 years. I'm not
totally up to date on everything, but I consider myself rather well
informed on the basics. As far as I'm concerned, Non-line-of-sight
(NLOS) is a marketing term and a theoretical exercise in reassembling
multipath signals into something useable. In my limited experience,
almost any NLOS system can be made to work for a while. The problem
is that the reflections and multipath move around as the objects
involved move around. The result is fatal un-reliability. What works
one day, fails the next. I can post MRTG graphs of signal strength
and S/N ratio of perhaps two NLOS links that I'm forced to deal with.
Huge random variations in both parameters along with signal loss for
extended periods are common.

>I'am a non native dutchy, so the right toplogy about somthing I have no
>knowledge of can be a bit lacking here. Sorry about that. I'll try my
>best.

No problem. Your English is much better than my Dutch.

>The distance between A and B1 is aprox 900 meters
>and from B2 to C is about 600 meters. There are no obstacles exept
>a lake between A and B1 (and a container terminal that doesn't reach
>above A and B, there is a lot of iron though, the containers.) This
>could influence the signal a little be I think.

Sounds like an ideal path. Both links are relatively short. 900
meters is nothing compared to the maximum 80km range of the Redline.
I'm a bit worried about why it has slowed down to 36Mbits/sec instead
of the maximum of 72Mbits/sec. At 0.6km, I would expect it to be
running at maximum speed. Lots of possible reasons including
interference, reflections, and improper setup.

>> Huh? Why a repeater? To what wireless device are they repeating to?
>> Methinks this is misconfigured or you're using the wrong buzzword.
>> Methinks both Redline boxes should be setup as a transparent bridge.
>
>The people who set it up told me that B1 is a repeater and also
>they named A1 as "office to repeater" and B2 "repeater to office"
>So they would be all transparent bridges? Makes much more sence indeed.

Well, they are correct that building B1-B2 is a repeater. B1 just
repeats what it gets from B2. Likewise, B2 just repeats what it gets
from B1. This is sometimes called a "back-to-back" repeater. It has
the advantage of being full duplex, where it can transmit at full
speed without any store and forward delays.

However, you stated that the radios were setup as repeaters. That's
not possible with either Redline or BR350 bridges. In order for this
system to work, *ALL* the radios must be setup as transparent bridges.
By transparent, I mean that it will pass more than one MAC address
allowing multiple computers to be bridged through the system.

>correct, they are bridges with a horrible web admin interface.

I've seen worse. Both the Redline and the Cisco have command line
interfaces. I've never played with the Redline version, but Cisco
BR350 uses IOS which are messy, but powerful. There are web
interfaces in both devices, but they are mostly front ends for the
command line interfaces.

>>>The ping results are good, 9ms peeks to 32ms. But when testing the speed by copying a large file from a fileserver, the download started at 30KB/s and then dropped to 8KB/s.
>>>Avarages at 12KB/s
>>
>> 9msec for such a system sucks. Something is wrong. Even the most
>> disgusting point to point wireless 802.11b link will do <3msec pings.
>> The Redline should do even less.
>
>Even worse, today (i was at the site for a half hour) the pings would go
>up to 1500, 3000 ms and back to 6ms. And other times stay just below
>50ms. Very weird situation.

Think about it. If there was a nice quiet link, with no interference,
the ping times would remain the same. However, if there was packet
loss anywhere in the system, it would manifest itself as wide
*VARIATIONS* in ping times. Traceroute should show what part of the
link is having problems. Again, I advise you to seperate testing of
the various parts of the system, and NOT test the entire system until
the component parts are proven reliable.

The large variations in ping time can be cause by packet loss anywhere
in the system. That includes at the interfaces to the bridges, the
coax mess in between bridges, NWAY failure, and wireless interference.
I'll bet on interference. It's common on tall buildings overlooking
areas where a large number of wireless users congregate. You might
look into the location of municiple wireless networks on 2.4 and
5.7GHz. Check:
http://muniwireless.com
for locations.
http://muniwireless.com/applications/737

>Good tip. But what if B1 or B2 is a faulty one. I would get bad
>ping results from A or C, because they have to go trough B1/2.
>A traceroute would workaround this, not?

You can ping from the other end of the link by logging into the
wireless router at the other end and pinging from there. It will be
slothish due to high packet loss, but the number will be real.
However, it might be best to do some travelling. 1.5km doesn't sound
like much distance to walk.

>Beter yet, the whole thing costs 10.000 euro's.
>I will call them tomorrow but have no fate in them fixing it.

That's $12,360 dollars. The hardware costs about:
RedLine AN-50 boxes and antennas are $3,455 each.
Cisco AIR-BR350-A-K9 bridges, plus antennas and coax,
are about $600 each.
for a total of about $8,100 list price plus incidentals.

I can see why they did not supply another pair of Redline routers.
The total cost for 4ea Redline boxes would have been $13,780, which
would have cost more than they bid on the project. There was no
shortage of Redline routers. They just replaced what was necessary
with whatever they had available and was cheap enough to make a
profit.

I don't want to get in the middle of a debate with your vendor. The
bottom line is that you paid a substantial amount of money for a
system that does not work even at the minimum theoretical level
(4Mbits/sec as limited by the 802.11b link). They also abandoned you
to do their troubleshooting for them. If you have some financial
leverage, I would use it.

>It's a backup for a fixed line whitch will arive within a month. We
>thought to bridge this perion with wifi. Sharing files with the
>fileserver on the other end would be great, but with this transfer rates
>this is not possible. Most important thing is stable connectivity with
>an AS400 application.

Well, it's certainly not stable with ping times varying all over the
place. I use the ping time (using pingprobe and MRTG) to measure the
system latency of my links. When they start to vary, I know I have
either excessive traffic or excessive packet loss.

File sharing is going to require some substantial bandwidth. It can
be done at the theoretical limited bandwidth of 4Mbits/sec for
802.11b. It would be better if it were running at the 10 times faster
speed of the Redline routers. In the US, if I want reliable operation
with no suprises, I spend the money on licensed wireless, with
coordination, to insure no interference.

>The company left all in default, the passwords still admin/admin for the
>redline's and the cisco's hace no authentication at all. Wonder how
>security is done here :)

What security? Defaults are just asking for trouble and hackers. You
will not find them on the 5.7GHz Redline link, but the 802.11b 2.4Ghz
link if full of hackers like me just waiting to pounce on an
unsuspecting wireless link and give themselves a tour of your system.

>The http and telnet interface of the cisco is so bad. I can reach them
>by telnet now and below a paste of their configs. Maybe some misconfig
>can be seen here.

I'll try, but I don't see any wireless (MAC layer) error statitistics.
See notes with <----- arrows.

>Redline A1:
>
>AN-50>show config
>sysname = kantoor to repeater
>sysdetails =
>ipaddr = 10.32.248.16
>ipmask = 255.255.255.0
>gateway = 10.32.248.1
>flowctrl = off <----- I don't know about this.
>eth = 100FD <----- 100baseTX FDX. No coax. Huh?
>http = on
>snmp = on <----- set the SNMP community names
>telnet = on <----- yech
>telnetport = 23
>rffreq = 5535
>autoscan = off
>dfsaction = 2
>antgain = 40
>txpower = 14
>atpcen = on
>adaptmod = on
>modreduct = 2
>ubrate = 36 <----- seems slow for short range.
>master = on
>encrypt = on <----- well, at least it's encrypted.
>encryptkey = ###
>llmod = off
>lmu = on
>ll = 0
>buzzer = off
>radio = on
>optionskey = ###
>flash 1: 1.32.003
>flash 2: 1.32.011 active
>
>AN-50>show stats
>macaddr = 00:09:02:00:38:A2
>swver = 1.32.011
>starttime = 0 day(s), 06 hr, 21 min, 04 sec
>rflink = Yes
>ubrate = 36
>rffreq = 5535
>txpower = 9
>cableattn = 7
>rfstatus = 0
>erxpkt = 54603
>erxpktd = 0
>etxpkt = 35250
>rssimin = -67.44
>rssimean = -66.69
>rssimax = -66.69
>sinadr = 20.64
>wrxpkt = 3195632
>wrxpktr = 1
>wrxpktd = 0
>wtxpkt = 3205747
>wtxpktr = 0
>wtxpktd = 0
>calcdst = 0.42
>hwrev = 3
>odu = T54


>Cisco 350 B2:
> Receive
> unicast pkts. 3745 39606 71609
> multicast pkts. 51743 0 229323
> total bytes 5525904 4282610 33605745
> errors 1 0 0
> discards 0 0 0
>
> Transmit
> unicast pkts. 3572 64618 73030
> multicast pkts. 32460 3 460987
> total bytes 3389515 12483286 118649149
> errors 1 682 0
> discards 0 0 0

There are a few errors (682) on the radio link show, but nothing
serious. I would think this would be a good and reliable link.
Hmmm...


>Event Log
> Time Severity Description
> 2 days, 06:31:41 (Info): Station [Magazijn]00409653ae3b Reassociated
> 2 days, 04:18:06 (Info): Station [Magazijn]00409653ae3b Reassociated

The above means the signal was lost for some reason and the bridge had
to reconect.

> 2 days, 04:18:06 (Info): Station [Magazijn]00409653ae3b Authenticated
> 2 days, 04:17:17 (Info): Deauthenticating [Magazijn]00409653ae3b, reason "Not
>Authenticated"

I think (not sure) that this is a WPA encryption key exchange. Not a
problem.

>Cisco 350 C1:
>not reachable at the moment

Why not?


--
Jeff Liebermann jeffl@comix.santa-cruz.ca.us
150 Felker St #D http://www.LearnByDestroying.com
Santa Cruz CA 95060 http://802.11junk.com
AE6KS 831-336-2558