Has anyone successfully deployed M5/AirMax products yet?

[whitedot]

I'm talking in a WISP/PtMP situation, with anything around 100 subscribers.

I'm looking at using NanoM5s as CPEs and RocketM5s with sector antennas as APs for a project where I live and am wondering if anyone is currently doing the same. I know the products are still new and the major firmware issues have only just been resolved but thought I'd ask anyway to get people's initial thoughts.

Please don't bother asking me more about my project because at present it doesn't formally exist and I know I have a ton of consulting and investigation to do... :icon_redface:

[whitedot]

I guess this is a no then. :icon_lol:

Anyone done any serious testing with the above setup?

[WifiMax]

I cant speak for everyone but from what i'm hearing on the iffy firmware... I would be terrified to hook up that many people with the current results people are posting with rockets/airmax

[whitedot]

Fair enough, I expected as much really. Just wondered if anyone had done much PtMP testing because the UBNT marketing of Airmax seems to be based on that yet most on here are just using them for PtP links from what I've seen.

[Airwip]

we just finished a new side with Airmax equipment.
Backhaul is made with 2 BulletM5HP (we may change that in future)
2 sektors with Rockets and 17db90° sektorantennas.
Router is a RB450G
we plan to load the sektors with customers in the next 4 weeks somewhere about 30-40 nanoM
each sektor. All test custommers are informed that this could run in problems .We are planing to ad clients step by step so hopefully we have a better chance to se if something going wrong at the time.
We just loaded all the equipment with the new firmware + double chek the hardware since a there where a few reports about bad hardware in the forum ..so far non of em had a problem.

I guess it's not what the initial poster was looking for but from my viewpiont a comercial wisp
must be nuts if he deploy a all new hardwaretype without being ultra carefully.

[vicenteroca]

I have a base station with 4 rocket + 4 sector 19db 90'.
I'm planning to start loading clients next week, and I am ****ing scared.
all I heard is problems problems problems, and stability issues. It's sad, because the new standard N and the mimo really show a huge huge improvement in speed, but, without stability this products worth nothing to me. And better not taking about the stock issues....

Right now, I'm considering switching to MK.

[whitedot]

Thanks for the feedback folks.

Let me be clear, this is months from being a reality so I'm certainly not on the verge of rolling anything out myself. Just wondered what you commercial guys were experiencing.

Not sounding great so far...

[whitedot]

One other thing I'd thought of btw was just to buy M kit and use it on 802.11a (once v5.1 is out and stable for that) for the short term. We'd be using legacy protocol speeds but that would be fine for now, and we can look into Airmax in the future without having to swap all the CPEs. The greater RAM and processing guts should help over standard NS5s, right?

[UBNT-Mike.Ford]

Thanks for the feedback folks.

Let me be clear, this is months from being a reality so I'm certainly not on the verge of rolling anything out myself. Just wondered what you commercial guys were experiencing.

Not sounding great so far...


Hello,

Please take into account that this is a support forum. People come here if they are having issues, generally not when things are working good. It represents a very small cross section of our customer base.

Thanks,

Mike

[UBNT-Mike.Ford]

One other thing I'd thought of btw was just to buy M kit and use it on 802.11a (once v5.1 is out and stable for that) for the short term. We'd be using legacy protocol speeds but that would be fine for now, and we can look into Airmax in the future without having to swap all the CPEs. The greater RAM and processing guts should help over standard NS5s, right?

Hello,
Yes it will perform better in an A enviroment as an AP over a standard NS5.

Thanks,

Mike

[WHT]

generally not when things are working good.

On the other hand, the first test of our 22 miles link was incredible!

Running 65 Mbps over 22 miles and costing under $300 was unheard of before last spring.

Ubiquity is rapidly becoming a ubiquitous name in unlicensed wireless projects.

When things don't work...its either a frustrated newbie, or a genuine firmware or manufacturing problem. The forum lets experienced users share their knowledge and point out problems that need to be fixed.

Just because you see so many problems and complaints, its because we have the forum to vent them out. Other manufactures have problems too, but you just don't seem them hung out on the laundry line.

[whitedot]

Hello,
Yes it will perform better in an A enviroment as an AP over a standard NS5.

Thanks,

Mike

Thanks for the confirmation, and I totally appreciate that people are coming here to get support. In fact despite the number of issues at present, I see it as a good thing that you guys are offering support so quickly and openly.

[rconaway]

As part of our Beta test, one of the locations is still running HD TV to 2 seperate clients. We had approval for another 10 or so when the cable TV client laid off most of it's employees and killed the project. 2 weeks ago they contacted me and the project is back on. The 2 clients are still running HDTV and I'm meeting with cable company next week to start planning more roll outs.

Although we haven't done as much high-density load testing for lots of users, I am confident enough to deploy the same 4 Rocket, 4 90 degree sector antennas setup on the City Hall at Sahuarita. It will immediately support 24 HD cameras (1600x1200, meaning at least 6Mbps per camera, 4 cameras per intersection). I have 360 degree coverage over several miles and plan on relaying up to 13 intersections over the next 6 months. This starts going live in 2 weeks as the intersections are completed.

I've designed similiar systems for City of Boulder City, City of North Las Vegas, City of Glendale (we hooked up all the traffice lights with control and cameras so the people coming here for the Super Bowl could leave quickly after the Cardinals won) using SkyPilot for the primary and Ubiquiti for some of the newer cluster areas. These are deployments that cost over $500,000 and are still growing. One area that has 5 Nano 5's on each camera is protecting a facility that costs $30,000,000. One of these systems has been written up in a couple magazines and has been reviewed by Homeland Security a couple of times.

I wouldn't be putting this system in with my name on it unless I was confident that it would work. The latest project is funded with federal dollars and if it didn't work, those are the boys I would have to answer to. Given that, I'm not saying the firmware is totally ready, but the reality is, it will be under development for a long time as every other product on the planet. The question is whether the product is stable enough to do what you need it to do right now. I'm good with it.

[WifiMax]

rory what kind of STB's do you use for the TV service if any?

[rconaway]

Haha, okay, what's an STB? Is that something I have to get a penecillin shot for?

Actually not that I'm an expert in that area yet but I believe they are Cisco's. Apparently they have a small buffer that helps with the wireless over some other products. We tested the system with MPEG-2 and MPEG-4.

[WHT]

<plonks Rory on the side of the head>

Set Top Box

[tfadgen]

WhiteDot, have you done the math??? 100 subscribers in how many days after release?? What do you have superinstallers and you are quick to deploy untested and unproven equipment??? My customers pay for reliable internet service, not to beta test wireless equipment??? My sincerest hope is to have a reliable AP/SU combo to cost effectively deploy. I started deploying Karlnet based AP1000's in early 2001, Trango Broadband in late 2002 till now(700 subs) and now I am looking for a reliable replacement in the sub 5.8 band. I have two test networks, one in 5.4 and one in the 5.8 band to test under noisey conditions. There are many things to test to reliability install and support internet subscribers.

[whitedot]

WhiteDot, have you done the math??? 100 subscribers in how many days after release?? What do you have superinstallers and you are quick to deploy untested and unproven equipment??? My customers pay for reliable internet service, not to beta test wireless equipment??? My sincerest hope is to have a reliable AP/SU combo to cost effectively deploy. I started deploying Karlnet based AP1000's in early 2001, Trango Broadband in late 2002 till now(700 subs) and now I am looking for a reliable replacement in the sub 5.8 band. I have two test networks, one in 5.4 and one in the 5.8 band to test under noisey conditions. There are many things to test to reliability install and support internet subscribers.

I appreciate the concern but if you read my posts in this thread you'll see I'm a country mile away from even formally starting this project, let alone rolling it out.

I was only asking. :icon_smile:

[whitedot]

Btw if the products are on the market I would expect them to be usable in a real world situation. Ubiquiti should be doing the beta testing prior to release, so I think it's strange that you suggest something that's freely available (well, when it's in stock) is not reliable enough to roll out.

I know there have been some serious firmware issues with the M products but it's weird that you imply that by thinking of using them I am doing anyone a disservice.

[WHT]

Btw if the products are on the market I would expect them to be usable in a real world situation..

Yahoo Weather Forecast for your Shirley McClain universe:
Mostly overly optimistic with scattered patches of surreality after first cup of organic coffee.


Edited to add...THAT'S A JOKE, DANG IT!

[whitedot]

:icon_lol:

Fair comment, but that's not my fault.

[rmeche]

Just thought I would add a good comment to this thread. We are an ISP in Texas, we have successfully deployed a Bullet M5 Backhaul link that supports around 30 clients. It is a 7 mile link and it has been online for 2 months. It has been 100% flawless. I have had zero percent downtime with the link. The speed is through the roof compared to the 8 meg link I replaced. I think Mike makes a good point, this support site is where people go that have problems and are trying to resolve them.

[kijoma]

hi,

I second that, we get people moaning they cant find much about us on broadband forum sites etc.. It is the same thing there, people usually only post when they have something to grumble about..

If its any help, the ethernet lockup problem i am informed was an Atheros issue and affected other users of the same chipset.

Ubiquiti look like they have fixed this problem which will no doubt benefit the other manufacturers as well.

I would give an example link from the MT forums on the same subject but i am IP banned :icon_mrgreen:

[whitedot]

Haha, why are you banned Bill?

Have you deployed any M products yet btw?

[lncommunications]

Hey this might be hole things up white...

http://www.ubnt.com/forum/showthread.php?t=14661

[whitedot]

Aye, been keeping an eye on that thread. :icon_confused:

[djtechman]

Let me preface, I'm a newbie to WISP but not to general RF or advanced IT/networking.

I've run some tests using Rocket M5's and 90deg/20dbi sectors.

So far I am not at all pleased with the range on the sectors.
First - I've run NSM5's in an AP/Station link (Nano to Nano) an achieved 79dbi+ at a distance of 5 miles... so I know my nanos are good & reliable for 5miles)

My initial tests involved a Rocket M5 + 90/20 sector on top of a mesa that sits about 365ft above my small town.
With approximately 1degree of actual downtilt (-1phys + 2 electronic = 1 actual), my trig calculations [(tan(89))*365=20910ft] are telling me I should be able to hit 3.96 miles in the center of the beam elevation IF the signal strength can hold out for that distance.
Initial testing (still not 100% confident that I got my downtilt right the first time) only got me about 1.5 miles max away from the sector before I was unable to associate with the AP. The test was run all the way across the horizontal beamwidth with the same results (even front-and-center at 3mi)

Anything over 2 miles gave me lower than -85dbi, and I'm confident in my aiming abilities because it was done at night with a tall tower's red beacon flashing directly behind my sector.

Re-testing this tonight and will be playing with downtilt to try and determine maximum range.

Unless I'm way off base on the above, It seems that the NSM5's are strong enough to achieve this distance, however the Sectors may be spreading me too thin to go the distance. This is underscored by the fact that, when I am able to associate with the R5, my nano's reported TRANSMIT signal level is always higher than the receive.

Any thoughts/suggestions are definitely welcome.

[popcorrin]

Are you running in N or A mode? Are all your pigtails tight?
You should actually be able to get more distance using the sectors since they have more gain than the nanos.

[spainuser]

Hi djtechman, Rocket M5 + sector 90º 20 dbi give me -69 to -70 dbm with NSM5 as client in 3.5 Km. CCQ 100 % and 130 Mbps link perfect. I set TX power to 27 dbm.

You must have any problem.


Thanks,


Juan

[skyhook]

First - I've run NSM5's in an AP/Station link (Nano to Nano) an achieved 79dbi+ at a distance of 5 miles... so I know my nanos are good & reliable for 5miles)

That TX Power?

[rconaway]

Your numbers may not be too far off but I think they are low by about 1-2 miles. I'm testing the sectors tomorrow with rockets tomorrow but it involves the parabolic dish on one side. I'll test the NS5M early next week though.

Those numbers are similiar to what I saw when I hooked some external sector antennas to an NS5. I didn't know at the time that there was a signficant signal loss in the external connector. I'm replacing them with bullet 5M's next week but I'm using the older hyperlink sectors. My target is 3 miles at -75dBm since I will pick up 5dBi.

[UBNT-Mike.Ford]

Your numbers may not be too far off but I think they are low by about 1-2 miles. I'm testing the sectors tomorrow with rockets tomorrow but it involves the parabolic dish on one side. I'll test the NS5M early next week though.

Those numbers are similiar to what I saw when I hooked some external sector antennas to an NS5. I didn't know at the time that there was a signficant signal loss in the external connector. I'm replacing them with bullet 5M's next week but I'm using the older hyperlink sectors. My target is 3 miles at -75dBm since I will pick up 5dBi.

Hey Rconoway,

Any update?

Thanks,

Mike

[rconaway]

No, couldn't get on the roof at City Hall. The janitor was too busy to walk 200' to open the door after I drove 2 hours and stayed overnight in Tucson and the IT and Traffic guys weren't there. Rescheduled for tuesday. I'm interested in seeing the NS5M and the parabolic dish numbers at that distance. I have some rockets going in with 23dBi Mars antennas on the traffic lights also at slightly longer distances.

[WHT]

Janitor must have been Civil Service.

[UBNT-Mike.Ford]

That sucks.

[WHT]

Since I'm east of the Continental Divide, the applicable vernacular would be, "that blows" instead of "that sucks".

[UBNT-Mike.Ford]

Since I'm east of the Continental Divide, the applicable vernacular would be, "that blows" instead of "that sucks".

ROFL good one

[webking87]

In regards to the topic of deploying Airmax equipment, I just recently finished installing three of the 5ghz Airmax 120 degree sector antennas along with rocket M5's on our 100 foot tower near San Antonio Texas.

So far I have about nine Nanostation M5's connected to the sectors, however airmax is currently disabled, as two of my clients had problems with latency while it was turned. (hopefully 5.1 fixes some of these things)

As far as signal strength on the clients, i haven't had great results. i'm currently getting results of between -75 to -86. the -86 signal is from a nanostation about 3 miles away, and the -75 is from a station about 1000 feet away. The stations with -75 to -79 have been working fine, but i'm really not able to push a lot of data to the nanostation 3 miles away (hopefully i'll be able to use the speedtest feature on the 5.1 firmware to see what kind of speeds i'm getting on these various locations)

[rconaway]

I don't understand the -75 at 1000 feet. That's a clue that something is wrong. You should be seeing -20 or some crazy number like that.

The -86 at 3 miles is probably slightly low. What is the power output setting on the radio?

[webking87]

Power at the nanostation M5 is at 20dbm, I increased the power to the full 27dbm, and am getting about a -70 to -71.

power on the rocket M5 is at 20dbm for testing now, which is still 4dbm over where it should be (assuming the 36 max less the 19 dbi gain of the antenna, then i should be at 16 as the max dbm) The only reason i have it higher right now is because i start to lose my connection with the unit 3 miles away if it's any lower...

:icon_rolleyes:

[pcjunky]

If you ssh into the units you will find iperf. Put the client in server mode and do the test from the AP.

[broken]

Your antennas aren't mounted correctly when you have such as low signal level. Did you remember that UBNT sectors have a electric tilt ?

[WHT]

Did you remember that UBNT sectors have a electric tilt ?I must have missed any reference to "electric downtilt" in the documentation.

Why does the top bracket have an adjustment for downtilt?

What if you don't want any downtilt? How would you "deactivate" the "electric tilt".

If you can explain to me how downtilt is achieved in an omni antenna then I'll know you understand how "electric" downtilting works. Actually, you don't have to explain it in depth... one key word would do it.

[george]

There is apparently 2* of electrical downtilt in the big sectors, and from what we've seen it actually behaves more like 4*. If you want a horizontal pattern you have to tilt the sectors up.

It shows clearly in the antenna patterns, check it out.

George

[whitedot]

I've used the M units now in many PTP links and they work well when tuned properly. As far as a PTMP setup they're basically useless until the ACK issues get fixed/removed with the new firmware. Otherwise you basically need all your CPEs the same distance from the AP as each other.

But soon they will be good. I have faith, maybe not much patience, but faith.

From reading this forum barely anyone is using these products in PtMP scenarios yet.

Which doesn't really help me. :icon_frown:

[rconaway]

I just finished my 4 rocket setup friday. I'm testing on Monday. I'll let you know how it goes.

[Ron]

I must have missed any reference to "electric downtilt" in the documentation.

Which means you either didn't read the documentation, didn't pay attention, or forgot.

The big sectors all have "Electric Downtilt 2 deg" and the small sectors have "Electric Downtilt 4 deg"

http://ubnt.com/downloads/AirMax5GSectors.pdf

WHT, please read before you type. I can't always be there to correct you ;)

webking87, do you know how sector antennas work and their radiation patterns?
It looks like you may have some aligned incorrectly.

[whitedot]

I just finished my 4 rocket setup friday. I'm testing on Monday. I'll let you know how it goes.

Is this the HDTV setup you mentioned previously?

Cheers. :smile5:

[rconaway]

No. Funny you should ask though. Those guys called yesterday and they want to expand the system later in January.

This is a new one to feed 50 HDTV video cameras to monitor 13 video cameras at intersections.

[webking87]

Hey Ron,

I failed to notice the 2 degrees of electrical downtilt that the sectors have.... i'll climb up in the next couple days and adjust the antennas to see what happens to the signal strengths.... Thanks!

[WHT]

Which means you either didn't read the documentation, didn't pay attention, or forgot.It was so small I didn't even see it.


I can't always be there to correct you ;)
I believe my esteemed colleague Rory pointed out how ridiculous your claims were.

[rconaway]

I also forgot to mention that the cable guys are running 3 HDTV clients off of a single AP and other than one TV hiccupping when someone changes channel, the still seem to be stable. This is with 5.0 firmware. The are upgrading to 5.02 to see what happens.

[kijoma]

From reading this forum barely anyone is using these products in PtMP scenarios yet.

Which doesn't really help me. :icon_frown:

We have two setup, have been for some weeks now..

The electrical downtilt is acheived by making the phase between the feed elements of the collinear array slightly short. it causes tilt and can increase sidelobes (well front lobes) due to the phase error..

it is usually frequency dependent so would be interested to know what the downtilt figures are for differing frequencies..

Of course Ubiquiti may of just mounted the array on a physical 2 deg tilt inside the case :icon_wink:

[davey]

Couple of washers on the top mounts should do it!

[davey]

And washers are not frequency dependent!

[whitedot]

We have two setup, have been for some weeks now..


How many users are you running on these PtMP setups?

[WHT]

And washers are not frequency dependent!Not so....the smaller the washer, the more frequently you'll drop it from a tower.

I'm just sayin'.

[rconaway]

You also forgot the fewer extra washers you have in that size is directly proportional to the chance you drop it and lose it.

[webking87]

Which means you either didn't read the documentation, didn't pay attention, or forgot.

The big sectors all have "Electric Downtilt 2 deg" and the small sectors have "Electric Downtilt 4 deg"

http://ubnt.com/downloads/AirMax5GSectors.pdf

WHT, please read before you type. I can't always be there to correct you ;)

webking87, do you know how sector antennas work and their radiation patterns?
It looks like you may have some aligned incorrectly.

Looks like actually tilting the sectors up some helped the problem. A few of the nanostations M5's that had been in the -78 range moved up to about a -70. Although the one that's 1000 feet away is still a -66... but that could be because of the angle going down from the 100 foot tower to the home at 20 feet up.....

[davey]

Not so....the smaller the washer, the more frequently you'll drop it from a tower.

I'm just sayin'.

Soooo true!

[UBNT-Mike.Ford]

Soooo true!


The smaller the washer, the more likely to drop it from a lab bench..as well...

Mike

[MDDA]

I'm talking in a WISP/PtMP situation, with anything around 100 subscribers.

I'm looking at using NanoM5s as CPEs and RocketM5s with sector antennas as APs for a project where I live and am wondering if anyone is currently doing the same. I know the products are still new and the major firmware issues have only just been resolved but thought I'd ask anyway to get people's initial thoughts.

Please don't bother asking me more about my project because at present it doesn't formally exist and I know I have a ton of consulting and investigation to do... :icon_redface:

Just completed phase 1 of a project doing this setup:
3 x rocket M5's attached to 3 x 120deg sectors
with NSm5's as cpe's

So far so good [os 5.0.2]
more details to follow.

BTW this project is in Manchester City Centre [UK]

Paul

[whitedot]

Nice one, I'm just outside Leeds. Used to work in Miles Platting if you know it.

What's the reason for doing something like this in the city centre?

[rconaway]

Update the firmware to the beta firmware. It works much better.

[davey]

Just completed phase 1 of a project doing this setup:
3 x rocket M5's attached to 3 x 120deg sectors
with NSm5's as cpe's

So far so good [os 5.0.2]
more details to follow.

BTW this project is in Manchester City Centre [UK]

Paul

I hear Ian Brown and Shaun Ryder are looking for broadband!

[samcamfilms]

No. Funny you should ask though. Those guys called yesterday and they want to expand the system later in January.

This is a new one to feed 50 HDTV video cameras to monitor 13 video cameras at intersections.

What are these cameras and resolution? 1080i / 720p ??

[rconaway]

We are planning on running them at 1600x1200 with H.264 and UDP. However, I just ran into an issue with a relay point that wasn't planned. I have to hit City Hall (sector antenna) from to a traffic light that will have 4 cameras and then relay from that point to another traffic light with 4 cameras. I would use rockets with the parabolics except that I have to go 10' above a traffic light hanging 20' over the intersection. I can't put 2 parabolics on that pole due to wind load and downright ugliness. Hard to explain why your brand new beautiful intersection now looks like a radar installation.

That means I need max bandwidth 4 miles back to the sector on Hall to a sector. Without the parabolics or grid parabolics, my next option is flat panel with a max gain of 22dBi. Subtract grounding, N-connectors, etc... and I'm down to 19dBi max. I have to run the channel width at 20MHz so I have some limits there also. That isn't going to cut it. I'm rethinking it right now.

[ciphercore]

We are planning on running them at 1600x1200 with H.264 and UDP. However, I just ran into an issue with a relay point that wasn't planned. I have to hit City Hall (sector antenna) from to a traffic light that will have 4 cameras and then relay from that point to another traffic light with 4 cameras. I would use rockets with the parabolics except that I have to go 10' above a traffic light hanging 20' over the intersection. I can't put 2 parabolics on that pole due to wind load and downright ugliness. Hard to explain why your brand new beautiful intersection now looks like a radar installation.

That means I need max bandwidth 4 miles back to the sector on Hall to a sector. Without the parabolics or grid parabolics, my next option is flat panel with a max gain of 22dBi. Subtract grounding, N-connectors, etc... and I'm down to 19dBi max. I have to run the channel width at 20MHz so I have some limits there also. That isn't going to cut it. I'm rethinking it right now.


I wouldn't mind seeing some pic's of these installs rconaway. or some network diagrams. Just sounds like an interesting setup.

[rconaway]

http://picasaweb.google.com/roryconaway/TownOfSahuarita?feat=directlink

[Ron]

my next option is flat panel with a max gain of 22dBi

There are panels which say 23dbi and 24dbi on the specs, from pacific wireless and arc.

Also, consider adding a panel where the sector is or putting a dish there.

[core]

I'm talking in a WISP/PtMP situation, with anything around 100 subscribers.

I'm looking at using NanoM5s as CPEs and RocketM5s with sector antennas as APs for a project where I live and am wondering if anyone is currently doing the same. I know the products are still new and the major firmware issues have only just been resolved but thought I'd ask anyway to get people's initial thoughts.

Please don't bother asking me more about my project because at present it doesn't formally exist and I know I have a ton of consulting and investigation to do... :icon_redface:

You can use M5's w/o any problem but first You must upgrade firmware, that comes with 5.0, to 5.0.2 for better device compatability and then waiting for version 5.1.

[rconaway]

I have to run dual-polarity antennas for the Rockets and I need maximum modulation. However, I haven't finished tested my parabolics to sectors yet. If the numbers get better, I have more options.

[sep78]

Where are you guys getting your ubnt dual pole sectional and dish antennas from ? I'm having a hell of a time finding any in the US and I'm seeing you guys have them installed in pictures already !

[rconaway]

I get mine from Microcom.

[sep78]

I checked their site, its showing no availability until Feb 2010. Is this because you purchased all their stock !?

[Ron]

A shipment came in about a month ago, some of us got some. I hear there is another shipment on a boat in the pacific at this very moment.

[adi_rotaru]

Hello
To be on-topic , maybe it helps to show a real-world throughput link based on 2 Rocket M5.
We are using 24 such ptp backhauls in our metro-network.
This one is using 2 sector 15 dbi antenna and is 1.5 km long.

[broken]

Adi are you using sectors for PtP Link ?

[adi_rotaru]

Adi are you using sectors for PtP Link ?
I don't get any omni's for the Rocket's :)) . Just a joke.
If the link is so short (max 1.5 km) the sectors are just perfect.
I also need some scalability so it can be easy transformed in a ptmp link .
For short links is also cheaper and as you can see: it works.
Bad part is that LAN interface is just 100 Mbps so all throughput tests are limited to the throughput of this interface. I have no doubts that it can perform faster .

[skyhigh]

Ubiquiti products are great as I was screwed over with Engenius cpe. After I switch my customers over to Ubiquiti NS2 the problems went away and I have to say that Ubiquiti has some very strong power cpe units. I'm now trying to switch my customer off the 2.4 GHz and on to the 5 Ghz M product from Ubiquiti. I hope that all goes well as I have just put up another tower as a back haul and is now starting to install the Rocket Dish and Rocket M5 for the link to the other towers of mine.

[rconaway]

Good move on your upgrade. Keep the following in mind, use the smallest width channels you need for your delivered bandwidth. For example, use 10MHz instead of 20MHz if you don't need 20 MHz. If you use 20MHz channels, keep 5MHz between the channels you use minimum, 10MHz if you can. If you use 10MHz channels, keep 5MHz between channels. Use the least amount of power you need to keep your CCQ at 100%. Use a tool like TheDude to monitor real-time bandwidth. Use the least amount of upload speed that meets your client expectations. If you are promising 3Mpbs upload, use 6Mbps on each client radio, don't use auto unless you need it. Also use G or A only if possible.

If you are using N radios, use the same rules modified for that speed. The exception would be the Airmax and AutoAck which should probably be on. I'm still testing to generate some basic rules.

[Ron]

If you are promising 3Mpbs upload, use 6Mbps on each client radio, don't use auto unless you need it.

Wait a sec. Are you suggesting limiting the wireless link data rate to 6 Mbps when the customer is expecting 3 Mbps TCP speed?

[WHT]

Well...yeah. Getting almost the advertised 3 Mbps is better than a fraction of that. Similar to if you subscribe for 10 Mbps DSL and the telco will consider 7 Mbps as acceptable.

But if you "promise" at least 3 Mbps in a service level agreement, then 6 Mbps isn't going to work.

[rconaway]

No. If you are promising 3Mbps up, then you probably need 9 or 12Mbps. Then I would use auto. My DSL lines in Phoenix and Tucson are lucky to hit 640K. Ron, you and I may not agree but give me more credit than that.

As for the higher bandwidths being more efficient. This is where you make the big bucks as the consultant. Let's take 2 scenarios.

1) All Ubiquiti equipment 802.11a equipment - Due to processor speed, Ubiquiti recommends a fixed setting on the AP side. This means you aren't getting the full efficiency of the band since the AP doesn't efficiently set it's AutoAck on a client-by-client basis. Now, if we could get CCQ% and ACK settings per client at the AP side, and in AP+WDS to AP+WDS, then we would have more information. However, with the current firmware which is generally very stable, it's still not quite smart enough and the processors are a little too slow to optimize the link quality to 100% to each user. Therefore, I make the choice to go for stability and sacrifice bandwidth efficiency. Now, with the M radios and the faster processors, I'm expecting that issue to go away and the radios in AP mode to be much smarter and fast enough to create efficient links bidirectionally. At that point, you set everyone to maximum throughput and let the radios figure it out. Note that in a Motorola Canopy or SkyPilot radio, you don't set an ACK. That's all done in firmware. Also keep in mind that the system, without polling, is a collision based system so depending on load, the slower speeds may end up being more efficient. If I'm wrong here, please let me know.

2) Ubiquiti Clients/Faster AP - If the processor was fast enough and the firmware in the AP was mature enough with the end result something on the level of SkyPilot or Motorola, then I agree with setting the clients to the fastest speed and letting the firmware figure it all out.

This is why $200 for a SkyPilot/Ubiquiti combination which has a maximum throughput of 12Mbps at the base station, may be a good idea if you want the stability and reasonable efficiency of the channel. It would definitely be easier to support for less experienced staff. I spend a lot of time working on getting CCQ% to 100% on a client by client basis which is where the firmware needs to go in an all Ubiquiti environment.

[Ron]

Using higher wireless rate is more spectrum efficient because it use less air time.

That's exactly right. Some don't realize that.

[Ron]

No. If you are promising 3Mbps up, then you probably need 9 or 12Mbps. Then I would use auto. My DSL lines in Phoenix and Tucson are lucky to hit 640K. Ron, you and I may not agree but give me more credit than that.

As for the higher bandwidths being more efficient. This is where you make the big bucks as the consultant. Let's take 2 scenarios.

1) All Ubiquiti equipment 802.11a equipment - Due to processor speed, Ubiquiti recommends a fixed setting on the AP side. This means you aren't getting the full efficiency of the band since the AP doesn't efficiently set it's AutoAck [...]

This isn't adding up for me. Are we talking about setting the wireless speed to Auto or setting the ACK to Auto?

Whether to use auto ack or manual ack depends on the situation and the product.

Whether to use a fixed wireless rate or auto, the answer is auto 99% of the time. In a wireless network, one of the worst things you can do is waste spectrum.

[Dave-D]

That's exactly right. Some don't realize that.

Ron, are you entirely serious? You think that a
channel can be re-used simply because it's idle
once in a while? Dave

[rconaway]

I was talking about the ACK setting. On the download speed, I set the AP's to 54Mbps and let connection work itself out. The problem comes in with the ACK setting. Mike suggested that as an AP, you should not use AutoACK. Right there, you have lost some efficiency. I have seen AutoACK work pretty well 99% of the time with the client transmit set to 6Mbps, 802.11g, auto-modulation off. I see on average about a 10-1 ratio or higher of download to upload. So calculating that out, I'm only losing about 8% efficiency on the wireless transport. Since my Internet bandwidth is 7.5Mbps/1.5Mbps at my fastest location shared with 150-250 people at anyone time, there is no realistic difference whether I use 6Mbps up or 54Mbps up.

Now, throw auto-modulation in with AutoACK on the client side, and then see how often the 3.5 firmware gets a little confused. After hours and hours of setting lots of radios, this is the combination that I have found makes for the most stable network and keeps CCQ at 100%. It also lets you use more Locos or in my case, Pico 2's which save me and my clients money. If the firmware ever hits the level of Motorola or Canopy in terms of stability and smarter ACK settings, then I would agree. The reality is that we shouldn't even be setting ACK or Modulation. The firmware needs to get to the level of figuring it out by itself.

[Ron]

Ron, are you entirely serious? You think that a
channel can be re-used simply because it's idle
once in a while? Dave


In 802.11a, most certainly. You didn't know?

You mean to say you never read the standard?
http://standards.ieee.org/getieee802/download/802.11a-1999.pdf

[Ron]

Since my Internet bandwidth is 7.5Mbps/1.5Mbps at my fastest location shared with 150-250 people at anyone time, there is no realistic difference whether I use 6Mbps up or 54Mbps up.

If the firmware ever hits the level of Motorola or Canopy in terms of stability and smarter ACK settings, then I would agree. The reality is that we shouldn't even be setting ACK or Modulation. The firmware needs to get to the level of figuring it out by itself.

You are right, in your case, there is "no realistic difference" in terms of what upstream speed you set. However, that is not the case for all WISPs.

One of our best selling points, when potential customers compare us to ADSL or Cable, is that we have symmetric speeds. People are willing to pay more for being able to send large photos faster, have higher quality video conferences, have more simultaneous VoIP phone calls, backup their data online and so on.

We have very few UBNT APs for 802.11a/b/g PtMP links. We use RouterOS as the AP. It's able to set a different ACK for every client, which I think is exactly what you are asking for. It's not polling, but it's as good as it gets without a proprietary protocol.

[rconaway]

I understand. Our back end is either T-1's or DSL circuits. In neither case do I have a lot of back end bandwidth. If I had more back end to deal with, I would have to work out the ACK settings at the base. However, with the N's I don't think it's going to matter anymore. The processor should be fast enought to handle the ACK issue efficiently as an AP. As the firmware gets more mature, that issue goes away. With polling on top of the faster processor, I think all those issues go away. I believe the firmware will get to the point of set auto everything and walk away.

[skyhook]

We have very few UBNT APs for 802.11a/b/g PtMP links. We use RouterOS as the AP. It's able to set a different ACK for every client, which I think is exactly what you are asking for. It's not polling, but it's as good as it gets without a proprietary protocol.

Hi Ron, how to set differennt ACK for every clients?
Dinamically or other via?

[Ron]

Hi Ron, how to set differennt ACK for every clients?
Dinamically or other via?

It's dynamic on the AP side. See attachment.

[Dave-D]

Ron, I have to admit I was being myopic.
I was thinking about the point-to-point
scenario I usually handle.

You bet: for a cluster of radios sharing
the same channel, faster is better. Dave

[Sirhc]

We have very few UBNT APs for 802.11a/b/g PtMP links. We use RouterOS as the AP. It's able to set a different ACK for every client, which I think is exactly what you are asking for. It's not polling, but it's as good as it gets without a proprietary protocol.

You would think that the Rockets could allow us to enter a default ACK as well as assign ACK's for each individual macs that will be associated with it and use the default if no entry is present for the associated mac.

And when set to auto ACK(when they get it to work right) it should be able to create an internal ACK table on the fly and keep track of individual associated macs and use the appropriate ACK values when transmitting to each client and the default ACK for things like beacons.

Just how does the UBNT deal with ACK from the AP to each Client when set to auto ACK. Does it keep some sort of ACK table?

[Sirhc]

Since my Internet bandwidth is 7.5Mbps/1.5Mbps at my fastest location shared with 150-250 people at anyone time

Your ****ting me here right? You share 7.5/1.5 with 150-250 people?

Holy balls Batman I wish I could get away with that! I have an OC12 feed that I pay for 100MB FD of use currently with a BAD *** cache server for less then 500 clients.

Last year I only paid for 20MB FD shared with 300+ clients and people *****ed and left me for cable. Now we are getting allot of people back since we offer faster service.

If it was not for the MIMO equipment that ubnt came out with at this price point I was going to give up on the residential market and concentrate on commercial PTP links and commercial high end links.

Everybody wants to use their internet now for hulu & netflix not to mention some also use vonage, broadvoice, or some other voip.

Some of my residential customers pull 100+ gigs per month and all I see is it getting worse as more and more people become net savy. I have some commercial accounts that pull close to a TB in a month! A freaking TB!!!!

I recently had a 65+ year old woman ask if we would help her set up a power HD antenna to her TV for local channels and her PS3 to watch movies from the net, and show her how to use an S-Video cable from her laptop to her TV for hulu. I explained to her that our squid cache server masks the identity of her PS3 browser so she can just use her PS3 to watch hulu. 65 years old! Like DAMN!:icon_eek:

[WHT]

Everybody wants to use their internet now for hulu & netflix not to mention some also use vonage, broadvoice, or some other voip.

Some of my residential customers pull 100+ gigs per month and all I see is it getting worse as more and more people become net savy. I have some commercial accounts that pull close to a TB in a month! A freaking TB!!!!

That's what happens when you don't have limits in your TOS and bandwidth management server.

We have a $24 a month account for 786 Kbps and 5 GB limit, then a $35 per month for 1,500 Kbps and 50 GB limit. If they don't like it, let them take their business to another provider. We make more money, by turning away some people. Business Economics 101 - Point of Diminishing Returns

[skyhook]

It's dynamic on the AP side. See attachment.
Yes, it's true... but allowing different ACK, can slow the system?

If you see status TAB of your WLAN, you can see a value of ACK (the max).
Are you sure thaht RouterOS don't use the max ACK for all clients?

[Ron]

That's what happens when you don't have limits in your TOS and bandwidth management server.

We have a $24 a month account for 786 Kbps and 5 GB limit, then a $35 per month for 1,500 Kbps and 50 GB limit. If they don't like it, let them take their business to another provider. We make more money, by turning away some people. Business Economics 101 - Point of Diminishing Returns

Chances are that Sirhc is in a populated area where his competitors offer higher bandwidth than you do, WHT, and Sirhc can also buy upstream bandwidth for much less than you do.

[Ron]

You would think that the Rockets could allow us to enter a default ACK as well as assign ACK's for each individual macs that will be associated with it and use the default if no entry is present for the associated mac.

And when set to auto ACK(when they get it to work right) it should be able to create an internal ACK table on the fly and keep track of individual associated macs and use the appropriate ACK values when transmitting to each client and the default ACK for things like beacons.

Just how does the UBNT deal with ACK from the AP to each Client when set to auto ACK. Does it keep some sort of ACK table?


Beacons don't have ACKs. That's why they are called "beacons." :)

Give it time. UBNT has been making software for less than 2 years.
There are still many things on the "wish lists".

Also, keep in mind, that if your links have high CCQ, which they should, then the ACK is not a huge deal.

[Ron]

Yes, it's true... but allowing different ACK, can slow the system?

If you see status TAB of your WLAN, you can see a value of ACK (the max).
Are you sure thaht RouterOS don't use the max ACK for all clients?

No, having different ACKs speeds up the wireless links.
CPU is cheap. Spectrum is expensive.

I'm reasonably sure.

Sounds like you have a routerboard. Why don't you test it our and report back?

[Sirhc]

Beacons don't have ACKs. That's why they are called "beacons." :)


You got me there, but you got my general concept!:icon_redface:

[skyhook]

No, having different ACKs speeds up the wireless links.
CPU is cheap. Spectrum is expensive.

I'm reasonably sure.

Sounds like you have a routerboard. Why don't you test it our and report back?

Yes, I use ROS.

At this time, I don't have any sure results, but only doubts for high ACK assigned dinamically.

I don't locate in ROS manual any info on this..

[rconaway]

Most of my areas have 2 choices for data, Satellite dial-up, or T-1's. In that case, they are so thrilled to have low latency, they are willing to be realistic on their use. I keep it to about 20-1. If they hit 20 users, I order a second T-1 and bond them.

In AP mode, version 3.5 carries different ACK's for each connected device. However, due to the speed of the processor under load, that slows down the radio. This won't be a problem with the M's due to faster processors but it's one of the limits of using the 802.11a radios as AP's. You just have to engineer with that thought in mind.

Tropos pioneered the idea of actually changing power output per packet to each user. I haven't tested that but I heard that it had limits due to CPU limitations also.

[whitedot]

Sounds like Bill is on the verge of deployment anyway, if I'm not mistaken. :smile: