3Com Building-to-Building Bridge

[Steve Sonntag]

We use a Win2000 server and all clients are Win2000 Pro. We use Terminal Services with Remote Desktop Protocol to access our accounting software. The accounting software is on the server at Building A and the TS users are at Building B.

We have a 3Com Building-To-Building Bridge setup. Both ends have a 50-foot antenna cable and 18db square panel antenna. The buildings are less than a quarter of a mile apart on different sides of the same four-lane highway. There is clear line of site between the two antennas except for some lines (maybe power, phone, CATV, etc.) on poles that run down the highway on both sides. The line of site is a few feet above the lines. I have configured them for 128-bit security.

The wireless system comes with the 3Com Wireless Bridge Manager software. It has a Remote Signal Strength Indicator. The RSSI generally shows an average of 74% (min 68, max 78) with occasional drops to min of 58%. I have no other diagnostic tools that I know of.

The problem is that the TS sessions have a tendency to drop at will and then take too long to reconnect for the impatient users. Which then snowballs into multiple TS sessions when the network reestablishes itself because the users are frantically clicking and rebooting. No errors are reported in any Event Viewers or the logs in the Wireless Manager. One or two days a week they have several dropouts, and other days is 100% uptime. And it does not seem to correlate with the weather.

My questions are:
-What is a GOOD signal strength that I should expect to see?
-Is there a relationship between RSSI percentages and data rate (1Mbps - 11Mbps)?
-How can I measure the current data rate?
-Is there a tool (inexpensive software) to monitor and record the data flow across the wireless link?
-What might be causing the dropouts?
-Would lowering the security to 40-bit or none help?
-Would we be better off using a different TS client instead of RDP (i.e. Terminal Services Client, Client Connection Manager)?
-What else am I missing that would be worth knowing or analyzing?

I have tried to ensure good power (UPS on all units) and good solid connections at all possible points of power, network and antenna connectors.

Steve

[Jerome Henry]

Hi Steve,

I'm not sure I can answer all your questions, but here are a few tips :

- There is no usual GOOD signal strength, but the more regular the signal is, the better will the communication become. Because :
- Yes, there is a direct relation between RSSI and rate. To make you understand how it works, imagine that each packet is sent to all directions. One will go straight to the other antenna, but some others will rebound on walls, widows or whatever else, and finally come to to the other antenna, with a slght delay due to the different distance factor, but still being the same packet !
The job of your antenna is to detect when a packet is redundant, and be able to get one packet after another. To do that, the best is to receive a packet, source quench (block the emitting antenna saying hey, wait a bit), wait for all the perturbations due to all the rebounds to stop, then receive next packet (I'm simplifying this process, called windowing). To make it short, let's say the more eprturbation you have, the lower RSSI you'll get, and the more time your antenna will have to wait to be sure that 2 packets won't come at same time, and the slower will your signal be...
- SO to jump to your dropouts, they are undoubtedly due to the fact that your antenna signal is not stable (moving from 70 % to 58 % as you say), and some packets are losts. If those packets are requests for replies, then your system deduces that the connection has timed out, and closes it.
- There are many tools that will confirm that point, check at :
http://bengross.com/wireless.html
- Don't turn to 40 bits or none, it won't fix things (packets are lost with or without encryption), I see so many networks I can hack without any effort, don't leave your doors open.
- Maybe is it possible to open a named pipe that would live the connection open for your TS, that is to say no timeout before shutdown (RDP-tcp properties, session part). You may also try to enlarge max simultanous connections (RDP-Tcp properties, network adapter part), so that if a connection is considered as idle, opening a new one won't get over max possible connections.
You may also start to audit TS events, to check if the trouble is really made by the antenna problem...

Tks to assign points if useful :]]

Good luck.

J

[Ron Kinner]

1. Per

http://support.3com.com/infodeli/tools/wireless/bridge/odb2bug.pdf

Page 21. 30 is good and over 40 is very good.

You might try a different channel in case you are getting interference or perhaps try horizontal polarization instead of vertical. Double check your alignment. Consider a taller tower. (You might experiment with telling it you are in a different country if you have that option. Other countries allow different channels and higher powers.)

Does the thing support SNMP? Easy to check the data rate then. If not what is it plugged into? Do any of the devices support SNMP? getif or mrtg can then be used to map the traffic.

What protocols are you using on the network? Is there a router on each end that could restrict traffic to just what needs to go over there or are you sending the whole mess over the link? Could be you are just getting some heavy network traffic that gets stuck in the wireless bottleneck. Do your PCs run netbeui or appletalk or anything but TCP/IP? Have you looked with a sniffer (www.snort.org has a nice one for free. It's actualy an intrusion detector but it works nicely as a sniffer.) to see what traffic is on the network that you don't need. Example: If you have Compaq PCs they all run several programs by default that talk all of the time even if no one is listening. The problem with a bridge is that it can't tell what should go over and what shouldn't so it usually sends everything. Your wireless link is the equivalent of a T1 or 1.5 Mbit/sec vs an Ethernet's 10 or 100 Mbit/sec so unless you are careful it is easy to overload it. It appears your bridge does have some discrimination in it since it talks about client lists and MACs so you might try throttling that down to the bare minimum but I can't see what it does for broadcasts. Seems like it would have to pass them. Perhaps you can set your switch to control the percentage of bandwidth used by broadcasts.

Getif has a reachability option which continually pings a target and graphs the resulting response. Might be useful.

http://www.wtcs.org/snmp4tpc/getif.htm

MRTG:

http://people.ee.ethz.ch/~oetiker/webtools/mrtg/

Ron

[Steve Sonntag]

- There is no usual GOOD signal strength, but the more regular the signal is, the better will the communication become.
Because :
- Yes, there is a direct relation between RSSI and rate. To make you understand how it works, imagine that each packet is sent to all directions. One will go straight to the other antenna, but some others will rebound on walls, widows or whatever else, and finally come to to the other antenna, with a slght delay due to the different distance factor, but still being the same packet !
The job of your antenna is to detect when a packet is redundant, and be able to get one packet after another. To do that, the best is to receive a packet, source quench (block the emitting antenna saying hey, wait a bit), wait for all the perturbations due to all the rebounds to stop, then receive next packet (I'm simplifying this process, called windowing). To make it short, let's say the more eprturbation you have, the lower RSSI you'll get, and the more time your antenna will have to wait to be sure that 2 packets won't come at same time, and the slower will your signal be...

Good Info Thanks

- SO to jump to your dropouts, they are undoubtedly due to the fact that your antenna signal is not stable (moving from 70 % to 58 % as you say), and some packets are losts. If those packets are requests for replies, then your system deduces that the connection has timed out, and closes it.
- There are many tools that will confirm that point, check at :
http://bengross.com/wireless.html

I'll have to check these out soon

- Don't turn to 40 bits or none, it won't fix things (packets are lost with or without encryption), I see so many networks I can hack without any effort, don't leave your doors open.

I really did want to avoid this

- Maybe is it possible to open a named pipe that would live the connection open for your TS, that is to say no timeout before shutdown (RDP-tcp properties, session part). You may also try to enlarge max simultanous connections (RDP-Tcp properties, network adapter part), so that if a connection is considered as idle, opening a new one won't get over max possible connections.
You may also start to audit TS events, to check if the trouble is really made by the antenna problem...

How would I create a named pipe or enlarge max simul conns? And on which device - the server, the client systems, or the bridges (haven't seen any parameters in configs for that on the bridges)?

--------------------------------------------------------------------------------
1. Per

http://support.3com.com/infodeli/tools/wireless/bridge/odb2bug.pdf

Page 21. 30 is good and over 40 is very good.

Sorry, wrong bridges here. Ours are indoors with antenna cables to the 18db antennas. The configs shown in this manual are much more extensive than with our bridges. 3CRWE90096A is our model

You might try a different channel in case you are getting interference or perhaps try horizontal polarization instead of vertical.

Maybe . . .

Double check your alignment.

Done

Consider a taller tower.

Only on simple mast

(You might experiment with telling it you are in a different country if you have that option. Other countries allow different channels and higher powers.)

No option for this

Does the thing support SNMP? Easy to check the data rate then.

No

If not what is it plugged into?

Simple Linksys switches at each end and one Win2k server and several Win2k Pro.

Do any of the devices support SNMP? getif or mrtg can then be used to map the traffic.

These are new tools that I'm not familiar with but will look into them.

What protocols are you using on the network?

TCP/IP and NetBEUI

Is there a router on each end that could restrict traffic to just what needs to go over there or are you sending the whole mess over the link?

No routers

Could be you are just getting some heavy network traffic that gets stuck in the wireless bottleneck. Do your PCs run netbeui or appletalk or anything but TCP/IP? Have you looked with a sniffer (www.snort.org has a nice one for free. It's actualy an intrusion detector but it works nicely as a sniffer.) to see what traffic is on the network that you don't need.

Does it record for periods of time, since the problem only kicks up every few days.

Example: If you have Compaq PCs they all run several programs by default that talk all of the time even if no one is listening. The problem with a bridge is that it can't tell what should go over and what shouldn't so it usually sends everything. Your wireless link is the equivalent of a T1 or 1.5 Mbit/sec vs an Ethernet's 10 or 100 Mbit/sec so unless you are careful it is easy to overload it.

Actually it can run at 11Mbps (counting the wireless overhead, 6Mbps of real data)

It appears your bridge does have some discrimination in it since it talks about client lists and MACs so you might try throttling that down to the bare minimum but I can't see what it does for broadcasts. Seems like it would have to pass them. Perhaps you can set your switch to control the percentage of bandwidth used by broadcasts.

Not this model

Getif has a reachability option which continually pings a target and graphs the resulting response. Might be useful.

Thanks to both of you for the info so far.
Steve

[Ron Kinner]

OK, You have the discontinued one. You did not mention if there were any errors in the log files or either bridge.

Do the drops in signal level coincide with the dropouts?

Ron