Wifi Hotspots using Bullet 2

[pcourtney]

we have been asked to blanket cover an area that is 7 miles average wide and 40 miles down a valley. Down one side of the valley 220 hotspots would be needed, and about 160 on the other side, we would like to use Bullet2's with 15dBi Omni antenna's for the hotspots, the roofs that we can use are all on top of public municipal buildings, so this is not a typical hotspot for a cafe or restaurant, but more for any tourist that happens to be in the vicinity using their laptop or PDA in their hotel, car park or just walking.

Anybody who uses the hotpsot is confronted with an HTML landing page, and when they enter the coupon code they get free access for 2 hours.

Luckily they have a data centre at one end of the valley, and the roof of the data centre is about 100ft high.

http://www.conniq.com/WiMAX/mesh-topology.htm

What would be the best way to provide the Bullets with IP from the data centre roof, have Ubiquiti any customers that have rolled out something similar to the diagram that Conniq use for a Mesh Network ?

NB They have fiber into the data centre from 2 telcos, and can have as much bandwidth as they want, several gigabytes could be available if required, so I don't see bandwidth being a problem.

[WHT]

How did you come up with the 220 hotspot APs to cover 280 square miles?
220 APs seems just a little bit low. Several thousand APs seems a bit closer.

http://forum.ubnt.com/forum/viewtopic.php?t=4808&highlight=

Chicago - 220 square miles, 7,500 access points, $18.5 Million or 34 APs per sq. mile

Google's Mountain View - 12 square miles, 400 access points or 33 APs per sq. mile

[UBNT-Mike.Ford]

Hello Pcoutrney,

I have to agree with WHT. The number of Hotspot AP seems very low. Also please be aware that the processing power of the Bullet's will limit the amount of units you can "mesh" using WDS.

Thanks,

Mike

[pcourtney]

I know what you guys are saying, I do - I do !

But my hands are a little bit tied, it's a govt funded project, it is probably nearer 300 sq miles anyway, but that does not matter, they own 220 buildings on one side of the valley, and 160 on the other side.

Each Bullet AP should with a decent omni should push out a signal
with a 1 to 3 mile radius, that would be fine to start things off.

The bigger question I guess I am asking, is if it is feasible to put
some Bullet 5's in the mix and mesh the whole network over the area ??

[pcourtney]

please be aware that the processing power of the Bullet's will limit the amount of units you can "mesh" using WDS


ummm, how many units can you mesh using WDS ??

I may have to place extra Bullets in the network to do this I guess !

[UBNT-Mike.Ford]

Hello Pcourtney,

For Meshing (having all AP's talking together) I would not recommend more then 6 hops. Every hop will kill your bandwidth by 50%

Thanks,

Mike

[pcourtney]

thanks Mike, that's what worries me, the far end of the valley is going to have next to nothing bandwidth wise, I may need to re-think this a little, perhaps it might be better if I have lots of small mesh networks getting their own feed via a Bullet 5 from the data centre.....

[pcourtney]

is it possible to connect the Cat5 on a Bullet 2 to the Cat5 on a Bullet 5 ?

that would be a cool way to get the bandwidth into a small mesh network of Bullet 2's !

Receive the IP on the Bullet 5 and pump it into the Bullet 2 - ummm !

[WHT]

You have a distinct advantage that is a government project. When it fails, no one will have to answer to any investors. I would not be so sure of myself if your calculation called for 220 and I suggested 330. But I'm suggestion over a magnitude greater.

Three miles? That's well in the radius that a commercial WISP can offer...with rooftop mounted end user points with high gain direction antennas and a 20 dB fade margin. But you are at street level with hand held devices. Cellular companies shoot for a 30 dB fade margin.

Take a look at this thread, http://forum.ubnt.com/forum/viewtopic.php?t=4808&highlight=

Ot of curiosity, what is the driving metric for this project? Did some government official come up with this 220 AP number? Did they quote a price and you figured that it would only pay for 220 radios?

[pcourtney]

You have a distinct advantage that is a government project. When it fails, no one will have to answer to any investors. I would not be so sure of myself if your calculation called for 220

It is not 220, it is 220 rooftops down one side of the valley, and 160 rooftops on the other side of the valley, this is a total of 380 wifi hotpsots, based purely on the fact that these rooftops are local govt. owned. We could place more AP's down both sides of the valley, but 380 is a good start.

[pcourtney]

Three miles? That's well in the radius that a commercial WISP can offer with rooftop mounted end user points with high gain direction antennas and a 20 dB fade margin. But you are at street level with hand held devices. Cellular companies shoot for a 30 dB fade margin

noted, but the intention at this stage is not to offer a perfect service, more likely they will be happy if most people in the vicinity can get a signal and if they have to walk towards the AP to get a stronger signal then that would probably suffice for a while.

[pcourtney]

what is the driving metric for this project? Did some government official come up with this 220 AP number? Did they quote a price and you figured that it would only pay for 220 radios?


No, it was purely based at this stage on how many buildings / rooftops they have in the valley, this gives them some wifi hotspots that can easily be operational at low cost, they have already organised for POE to be available on every rooftop, and modified the roofs of about 135 buildings so that it is easier for an omni directional antenna to be placed in such a way that it is not being overshadowed by other buildings, many of the rooftops were not flat, and so access under the roof was provided for, so an antenna could poke out of the top of the roof, the work has taken over 6 months, and we are now finalising who our vendor of choice should be.

[pcourtney]

is it possible to connect the Cat5 on a Bullet 2 to the Cat5 on a Bullet 5 ?

that would be a cool way to get the bandwidth into a small mesh network of Bullet 2's !

Receive the IP on the Bullet 5 and pump it into the Bullet 2 - ummm !


thinking about this some more, we still need POE to these Bullets :(

[UBNT-Mike.Ford]

Hey Guys,

You can definitely go B2 <--ethernet-->B5

Thanks,

Mike

[CDX825]

Would have to be set up as follows
B2-ethernet-poe injector-ethernet-poe injector-ethernet-B5
The cable between the two injectors would have to be a cross over cable right?

[WHT]

http://www.meshdynamics.com/third_generation.html

http://www.meshdynamics.com/images/MDCompetingArchitectures.gif

[samcamfilms]

Wow looks like a big project.

Why don't you just "group" say like 50 ap's to a 20-50mbit + back haul link.

Dose this organization have access to fiber in the region? think outside the box... mesh might not be the way to go in this case..

EDIT= yes mesh is good, but the bandwidth would be better if linked via fibre

[pcourtney]

WHT, very interesting, do you use any of their kit ?

http://www.meshdynamics.com/third_generation.html

[pcourtney]

Why don't you just "group" say like 50 ap's to a 20-50mbit + back haul link

earlier in the thread I was thinking that myself :-)


does this organization have access to fiber in the region?

No they don't I am afraid.

[WHT]

No...we don't use the Mesh Dynamics, but its on a A List of things to look at this spring.

[pcourtney]

Mike, to look at duplicating, 3 radios, 4 radios, or working with them ??

[kijoma]

hi,
it wouldnt need to be cross as the units are auto select (never sure how two auto selects decide who is going to be who :) ).

Just got in this thread, to me you need a 5 GHz backbone pointing down the valley and use PS5 or NS5 to pick it up and do as you suggest, poke it into the local Bullet2's

to save PoE adaptor headaches and having so many crimped plugs/wipe blade sockets in line threatening you with nasty support visits (the bullet doesn't come with a PoE box anyway), i would take the cable that goes between the two units, put the middle of it (or any point along its length) into a box via glands and split open the outer and hard solder the power feed to the blue + / brown - wire pairs.

To do it cheaper, forget the box, do the same then us amalgamating tape to seal up each joint and then the total three way cable join. then run the two core or bit of cat5 you used for power down to your power adaptor (suggest a 18V reg one, min 0.8 Amp as it is powering two radios).

[pcourtney]

Hi Bill,

we already use CHOCBOX 2amp 8 pole boxes, they are only

http://www.getplc.com/news/news019.asp


http://www.chocbox.info/capacities.html

under a quid from Screwfix

http://www.screwfix.com/prods/46538/Electrical/Cable-Accessories/
Cable-Connectors/Chocbox

[pcourtney]

Hey Guys, You can definitely go B2 <--ethernet-->B5 Thanks, Mike

because both the B2 and B5 can auto detect whether the cable being used is straight through on both Bullets, with both cat5e cables wired into a plastic junction box like the Chocbox - no need for POE special units ?

that's clever !


http://www.chocbox.info/techspecs.html

[WHT]

We use a lot of this Radio Shack barrier strip and outdoor rated grey PVC electrical junction boxes.

http://telecomanddata.com/junction_box.jpg

http://www.radioshack.com/product/index.jsp?productId=2103232

http://rsk.imageg.net/graphics/product_images/pRS1C-2265725w345.jpg

[laser3]

Hi pcourtney sorry it took so long for you to get a Difinitive answer, which i believe Bill answered correctly you best bet would be to have a sort or dual mesh 5Ghz running the Back Bone & 2.4 Ghz running the public mesh I have also noticed you in the UK. it sounds like a very interesting project & if you don't mind I would like to work closely with you on this, you can email me on mysmallspace@hotmail.co.uk or skype me (lasermike33) I just got a new mobile & I don't know the number :lol:

[kijoma]

Hey Guys, You can definitely go B2 <--ethernet-->B5 Thanks, Mike

because both the B2 and B5 can auto detect whether the cable being used is straight through on both Bullets, with both cat5e cables wired into a plastic junction box like the Chocbox - no need for POE special units ?

that's clever !


http://www.chocbox.info/techspecs.html

hi, watch out for choc block connectors though, they screw down onto the wire and as it is thin/solid core it breaks pretty easy in that setup.

if you push the folded wire into a gland in the box then you only need to break out the power pairs. The green/orange data pairs remain unbroken between radios.

I am all for removing unnecessary crimp connections and wipe blade plug/socket setups with external kit :) , one bit of damp and the connection integrity is gone!

[rconaway]

I'm going to put my 2 cents in at this point. Basically it sounds like you have to provide WiFi around 380 buildings. As WHT said, no way. If you only want a few hundred feet from each building, then maybe. If you put NS2's on the outside of the buildings with LOS to the street, you might get about 1500' or more.

Cluster the buildings into groups that have LOS within 3 miles or so (to keep backhaul modulation high). Find a center backhaul, VPN wired or wireless and use an expanding star concept. Use the PS2's to create the cluster backhaul and then parallel the backhaul with the area coverage radios, 2.4GHz. Use 5GHz backhauls to a start branching design with WDS. It's going to take a lot of driving around to get this mapped out correctly.

I just did one similiar to this that required 20Mbps at the street level in a 5 square mile area. Assuming that the clients have NS2's at a worst case scenario, they may have to go 10' off the ground. I am using 30 AP locations with multiple AP's at every locations. It's kind of like guerilla warfare, you latch onto every rooftop, every pole, etc... that you can get on. That took 2 days of driving around the areas mapping out the design.

[pcourtney]

thanks

>Basically it sounds like you have to provide WiFi around 380 buildings.
>if you put NS2's on the outside of the buildings with LOS to the street,
>you might get about 1500' or more

that is all that has been asked for at this moment in time :-)

Simple Wifi hotspots out in the street around the 380 public buildings they have in the area, swimming pools, libraries, schools etc.

We are now looking at clustering these 380 sites into groups, and then figure out the best way to backhaul using B5's

[WHT]

"that is all that has been asked for at this moment in time"

Translation: They will most certainly want more, necessitating you to completely redesign your infrastructure to the point of being easier to start over from scratch. :lol:

[rconaway]

WHT is correct. I would suggest that you get a project scope and statement of work in writing before you start. I see this project heading for a difference of opinion but not knowing your clients, it's hard to assess.

[laser3]

Like i said if you need any help with planning your deployment field test and things of that nature drop me a line mysmallspace@hotmail.co.uk I would be more than willing to help.

[sankofa]

I am not sure when the RouterStation will be available, however IMHO I would build a robust mesh network with 3 radio nodes. 2 - 5Ghz and 1- 2.4Ghz. It should cost you under $300.00 per node. Knowing these city deployments they start of with internet access, then the want cameras etc, etc. Even if they do not add-on other services you will have a more solid network. Just my 2 cents. Good Luck.

[brasileottanta]

Hello Pcourtney,

For Meshing (having all AP's talking together) I would not recommend more then 6 hops. Every hop will kill your bandwidth by 50%

Thanks,

Mike

But , this is not a really mesh-network . In pure mesh-network , no bandwidth reduction .


Marco

[UBNT-Mike.Ford]

Hello Marco,

Pure mesh uses dual radio's and self healing networks to achieve this and the reason behind the cost of 2k to 8k US dollars per node. Our radios are less then 100 dollars.

Thanks

MIke

[brasileottanta]

Hello Marco,

Pure mesh uses dual radio's and self healing networks to achieve this and the reason behind the cost of 2k to 8k US dollars per node. Our radios are less then 100 dollars.

Thanks

MIke

Hello Mike !!!


Yes , I know !!!!! Our network is Ubiquiti based on CPE side. But for BS we use MT ( actually ... )

[rconaway]

Bandwidth reduction is not a matter of mesh, it's a matter of how many radios are handling the hop. I'm generalizing as some vendors have instituted proprietary methods with single radios.

The least expensive mesh radios, pepwave for example, use a single radio and you get a general loss of at least 1/2 your bandwidth on every hop. Move up a little to a Proxim whichs costs about $1000 and uses a 5.8GHz radio for backhaul and a 2.4GHz AP radio, and the bandwidth loss is the same, you just aren't asking a backhaul radio to handle AP mode simultaneously. Move up further to a Bel Air which might use 2-4 5.8GHz radios for backhaul and the costs jump to as high as $10K or more and you probably don't have any appreciable loss in the backhaul.

Yes, every vendor has several models and there are high-end versions of Tropos single radios, etc..., I'm being very general about everyone's first versions.

Given that, the best solution with the most throughput in my opinion, not withstanding any issues that have seem to crop up recently, for an almost all Ubiquiti AP would be 2 NS5 Locos with a Picostation 2 with 15dBi antenna and a 5-port switch. Put all the power cords and switch in a Nema enclosure (Home Depot $25.00) and you have a $400 AP that rivals $3000-$8000 mesh radios. Make a Nanostation 4.9GHz radio that has AES, WDS, and full FCC certification and I wouldn't be able to keep them on the shelves. Price it at $299 retail with a wholesale cost of $199 (so I can make a profit on the hardware). I would be sending you Christmas cards the rest of my life.

[brasileottanta]

Bandwidth reduction is not a matter of mesh, it's a matter of how many radios are handling the hop. I'm generalizing as some vendors have instituted proprietary methods with single radios.

The least expensive mesh radios, pepwave for example, use a single radio and you get a general loss of at least 1/2 your bandwidth on every hop. Move up a little to a Proxim whichs costs about $1000 and uses a 5.8GHz radio for backhaul and a 2.4GHz AP radio, and the bandwidth loss is the same, you just aren't asking a backhaul radio to handle AP mode simultaneously. Move up further to a Bel Air which might use 2-4 5.8GHz radios for backhaul and the costs jump to as high as $10K or more and you probably don't have any appreciable loss in the backhaul.

Yes, every vendor has several models and there are high-end versions of Tropos single radios, etc..., I'm being very general about everyone's first versions.

Given that, the best solution with the most throughput in my opinion, not withstanding any issues that have seem to crop up recently, for an almost all Ubiquiti AP would be 2 NS5 Locos with a Picostation 2 with 15dBi antenna and a 5-port switch. Put all the power cords and switch in a Nema enclosure (Home Depot $25.00) and you have a $400 AP that rivals $3000-$8000 mesh radios. Make a Nanostation 4.9GHz radio that has AES, WDS, and full FCC certification and I wouldn't be able to keep them on the shelves. Price it at $299 retail with a wholesale cost of $199 (so I can make a profit on the hardware). I would be sending you Christmas cards the rest of my life.

Hi :-)

this is one solution. There are more . Depend on scenario and what you want .

In my country isn't possible use 4.9ghz band. So , we have other solution to risolve the 1/2 bandwidh problem.

We have a p-t-p at 5.3 ghz at 40mhz channel for 108Mbit in every basestation using PS5 ( only for link ), BS ( MT ) at 5.5 ghz for residential customer and 2.4 for public hot-spot. Gateway is connected to a fiber at 100Mbit . This is similar to BelAir ( with better performance ) , but less cost.

http://www.gnc.it/images/Foto-BTS%20004.jpg
http://www.gnc.it/images/Foto-BTS%20034.jpg
http://www.gnc.it/images/Foto-BTS%20035.jpg


Marco

[rconaway]

We can't use 4.9GHz for public, 5.3 and 5.8 only. 4.9GHz in the US is only for public safety. Unfortunately, there are very few products there with FCC certification. I was just making a "suggestion".

I wouldn't say that it has higher performance. To be more accurate, let's just say that it's within 25% of the performance in 5.8GHz thoughput, which given the price difference, is extremly impressive.

[brasileottanta]

We can't use 4.9GHz for public, 5.3 and 5.8 only. 4.9GHz in the US is only for public safety. Unfortunately, there are very few products there with FCC certification. I was just making a "suggestion".

I wouldn't say that it has higher performance. To be more accurate, let's just say that it's within 25% of the performance in 5.8GHz thoughput, which given the price difference, is extremly impressive.

In Europe isn't possible every use of 4.9Ghz band. But , we have 19 non-overlapping channel in 5.0Ghz :-)

And we have Hiperlan/2 , more different from 802.11a , but no vendor have put developing in this standard.

Hiperlan/2 is ,de facto , wimax in unlicensed band. TDMA and TDD .

:-)

Marco

p.s. for more info on Hiperlan/2 take a look at


http://www2.ing.unipi.it/~d8328/documenti/wlan/Data/refer/stdhiper/ts_10176101v010201p.pdf

[WHT]

That better than the six 20 Meg wide non-over lap we have in the U.S.