What is “the” way to setup a Sector based tower ,

Would be nice to figure this out for once,

At the moment we have several towers with 3x120degr sectors in top.

The setup we are using is a kind of experimental one, just to see if roaming/same SSID’s on all towers works.

Setup router ;

- All 3 2,4Ghz sector radios into same bridge
- Sectors 2,4Ghz onto 1/6/11 channels
- All sectors same SSID (ssid broadcasting disabled)
- Sectors ( http://www.stelladoradus.com) are around a pole back to back with a 20cm between.

All testing etc did go well, but this weekend after installing a client nearby a tower things got crazy on that tower ;

There are a 8 (WRT54G / CB3) clients with external antenna on this tower and working ok, but when i power up client 9, who is a senao/enginius CB3 nearby the tower without any external antenna, the tower goes crazy…

All wireless clients suffer signal lost/ connection drops then etc

Don’t know exactly what courses the pain, but likely does have something to do with “Hidden node effect “ ? or radio interference, would a high gain antenna solve this ?
Or is there a way around to hook up the client Mac to a one sector only.

I hope to open a discussion / gathering experiences with a “how to” at the end for everyone how needs to know how to build a good/solid tower with sectors.

We have this exact setup on 5 towers... SSID broadcast off / same SSID, different channels... We have clients hop from sector to sector, tower to tower all the time. We have a few mobile clients that go all over the place, and others end up hopping, let's say if there's an equipment problem or too much interference.

I would look at a few things:

1. I set all my CB3 clients to RTS threshold 512 or less - this helps hidden node A LOT. You may even consider going as low as 256 or 128. There's nothing to set on the tower, FYI.
2. Maybe a bad CB3 on #9 - I can't see why it would knock out the WHOLE tower unless it's got a virus or something on it - interference would knock out a radio but not a whole tower, esp. on different channels.

Last night i switched on the #9 CB3 again and put its Mac (and other connected clients) in the wireless tables  access list, so clients stay with the sector the were first connected to for as good as I know, and everything seems to be stable now, but we don’t know for sure in 10 minutes, so we wait and see.

Me think the problem is probably that the CB3 client nearby the tower (it’s a 79mW) is roaming like a mad man between the sectors and is eating up almost all the airtime ? (didn’t checked the log btw)

Also found another (basic) problem at that tower, we did had a LOT of rain here the last week and now one sector doesn’t pick up ANY clients anymore, so that’s what has to be repaired tonight first, probably a water problem…never going to use those crimp connectors again with that useless tape, its always somehow going to leak or something, best connectors we use are the (Gandalf) H500 golden connectors, never have to tape them because the construction is weather proof ( just a tip)

Quote<directwireless>
1. I set all my CB3 clients to RTS threshold 512 or less - this helps hidden node A LOT. You may even consider going as low as 256 or 128. There's nothing to set on the tower, FYI.

What exactly does the threshold do ?
And does it have (side)effects on traffic/speed etc ?

Oh, water leakage has been a headache for us too - we had some guys climb one of our taller towers to put up equipment, and found months later he didn't put any tape on the LMR400 connections (crimped also).

I can't see why it would keep roaming - I have a hard enough time getting them to switch when the signal goes away.

Setting the threshold does this... Anytime the client goes to talk to the tower with a bigger packet than the threshold, it says "Hey, Can I talk?" (Request to Send) and the tower says "Sure, no problem." (Clear to Send). When the other clients hear the CTS, they know not to talk for a random timeout (in microseconds) and then they try again. Then the hidden node problem is mostly solved. Smaller packets, are sent without a request to send, which is a performance tradeoff, since it is less efficient to ask than to just keep sending it.

CB3's usually have a feature called AP Density, which if you set to low, should make it switch less often.

I have a couple of questions along this line. I am going to be changing over to sectors which means my current clients must be moved. Clients are not running MTK but the sectors will be. I assume you are setting up the MTK as ap-bridge. Therefore you must route traffic to your clients through the appropriate sectors. Which in my mind means each sector must have its own subnet to be able to route properly. So I assume you must see which sector the client is associates to then assign them an ip within that sector's subnet. So why use the same ssid on the three sectors? Is there a better way to do this? From reading other threads, bridging is not an option due to the differences in ap and client software, routing issues and I thought with the same ssid I would have association problems between adjacent sectors. Also, bridging leads to its own problems such as loss of some functions you may want on the that sector for diagnostics or control that you have in ap-bridge mode. I tried bridging on a previous tower but found that I couldn't connect to it from my home, only could connect if I was plugged into the switch at the base of the tower ie on the same subnet as the ether ports.
Thoughts?

tk

Bridging can be a headache sometimes, but for us it's paid off in redundancy. As an example, we had a AP-sector go offline on us, and another panel on the tower took over the clients almost instantly. Routing would not have allowed that and they would be offline until it was repaired.

I agree.. get briding right for redundacy and you can go forever as long as you replace the hardware that screwed around.

I have APs with 4 x 90 sector for close to 2 years with no hassles at all, even with an occasioanl water leakage problem.

Hey thanks for the insight from real WISP operators -
Can I be nosy and ask:

Do you bridge all 4 wireless interfaces to the ethernet interface?
Any problems with this (P2P, viruses etc)?
Do you use WDS?

Would appreciate to know -

Regards -

Thank you very much for sharing the advice -

Agreed, RSTP is a good solution for resilience.
Isn't it possible to do something like this within MT though? I'm not personally familiar, but some of the L3 protocols should be able to do that I guess.

Regards

I tried MT with STP, setup to prioritize 2 backhauls, and I had more headaches than I can count... I had things crashing on opposite ends of the city one after another from it taking too long to find network loops.

Yes I can believe that. I notice MT has STP built in, but the protocol takes a long time to "converge". I think it's used to avoid loops in mostly-static networks, not for rapid fail-over resilience.

RSTP converges in seconds or under, and though there are limitations in the number of hops (5 I think, someone correct me if I'm wrong) it means traffic doesn't suffer much.
I know some resellers use it to provide fail-over with laser links back to RF backup etc.

I wonder if MT could put RSTP in the router?

Regards

Directwireless,

I am interested in the configuration of you tower with 3 sectorial antennas and roaming facility for mobile voip handset. Do you have any experiences with that ?

thanks

Gianluca

As you told that you on 9th CPE and you face packet loss issue,The same problem i had faced, I am using 2 AP, Omnitik and 120 degree mantbox sector, As one of my user access pirated software website using proxy all cpe which is connected to the same ap more than 78 signal on ap side suddenly face packet loss issue. then i block that cpe and everything is fine.