Ask the Experts: 3 WISPs from North America May 20th


@Ryo Haines wrote:

Good to know, I'm pretty remote as well and there is only 1 other WISP in the area.  Fortunately, he is strictly Cambium gear and is restricted to 5.7/5.8 so the rest of the spectrum is clear for me.

It would be great if vendors would have a loaner or something to test with so we could get the best link depending on each situation.


you can cheat a little.   using a conectorized radio, you can install 1 pair of dishes and run each brand for a week off of the same dish, compair your finding and make a choice, its not exactly a loner, but you could always return the unused radios.    

one more thought about the mimosa, we haven't used them because they need so much spectrum to get their advertised speeds. using a smaller space, they don't perform any better then the AF5 and the AF5 is a full duplex radio and has awesome front to back ratios and sync. they run about $400 more per link, but given the specs and how far apart the FTB is we won't use them for spectrum efficiency reasons.      I think the mimosa has a few great application points, but i'd rather spend the little extra for the bigger result. 

1 Like

@Chris_Bay wrote:
using a smaller space, they don't perform any better then the AF5 and the AF5 is a full duplex radio and has awesome front to back ratios and sync. they run about $400 more per link, but given the specs and how far apart the FTB is we won't use them for spectrum efficiency reasons.

One issue with the AF5 is that they have two different physical models to choose from depending on what 5GHz band you want to run on... unlike many other 5GHz radios that support UNII-1, 2, 3.

Also, I've heard, and maybe Chris can comment on this... that the AF5 is a full-duty radio, meaning it's TX'ing constantly, even when it's not transmitting data. I've heard reports that the AF5 will disrupt other 5GHz radios on a tower if the radios are anywhere near the frequency of the AF5.

Great feedback, thanks everyone...

I'm really intrigued now about a new CNS server, I guess bigger and better things are to come for sure.

Maybe one of these days in your spare time (I'm sure its rare and far between) but maybe one of you could hold your own webinar?  I attend many of the Cambium ones and they are pretty informative, but it would be great to see an environment live and working and the results you are getting with the gear.  Everything is speculation or just a good guess until its up in the air and connected.

I always find it to be real good reality check to see other environments and see what you can take back to improve your own.  I think we would all benefit from that.

Just a thought and thanks again for your time today,

Ryo

1 Like

Our guests will be back in a few hours at 6PM CST. You can post your questions and comments about anything here and they will get to your questions.


@Eric Ozrelic wrote:

@Chris_Bay wrote:
using a smaller space, they don't perform any better then the AF5 and the AF5 is a full duplex radio and has awesome front to back ratios and sync. they run about $400 more per link, but given the specs and how far apart the FTB is we won't use them for spectrum efficiency reasons.

One issue with the AF5 is that they have two different physical models to choose from depending on what 5GHz band you want to run on... unlike many other 5GHz radios that support UNII-1, 2, 3.

Also, I've heard, and maybe Chris can comment on this... that the AF5 is a full-duty radio, meaning it's TX'ing constantly, even when it's not transmitting data. I've heard reports that the AF5 will disrupt other 5GHz radios on a tower if the radios are anywhere near the frequency of the AF5.


good point eric, the AF5 does not use the entire band UNII 2, and no UNII-1 support.

I'm not sure if its a truely full duty radio or not, however we do treat it as such, because of the loading we put into them.    our smallest spacing to the face of the AF5 with an EPMP is 15 mhz without issue (uni-II-2 band)  i haven't tried the upper segament yet.    we do try and keep our backhauls as low as possible on a tower away from the multi-point gear.


How is your network setup? Is your network bridged or fully routed? How much is interference a problem for your network?

Thanks,

Sanay


I thought I would add a little to this as well. We recently moved from a bridged to a fully routed network using MikroTiks and OSPF. Our CPEs are configured as bridges and the customers are given public IPs for their routers specific to the tower they are connected to (we use /28-/24 blocks depending on the need and add ranges as necessary). The MikroTiks at the tower act as the public gateway for those customers and then route between the towers/off network from there.

Most interference we see in network is self-induced. We started with 900 and 2.4 and saturated our core network, and added a layer of 3.65 and now ePMP 2/5Ghz. Our core network was built in an area that was undeveloped, and we are now getting into areas with competition involved. In those areas, most operators are in the 3.65 or 2.4 band, and are willing to work with each other to channel plan in order to keep from either one of us hindering the other's frequency use. 


@Ryo Haines wrote:

Hello everyone,

How do you successfully manage your environment?

I'm coming from a Ubiquiti / Aircontrol and slowly migrating to a Cambium / CNS world and have run into some shortfalls in managing the environment.

For example:

  • I want alarms in place when a backhaul goes down or an AP's utilization goes high.  (I've started using PRTG for this)
  • I want to see an overall view of my environment without having to drill down to every device.  (aircontrol works pretty good for this by sorting)
  • How do you know what devices are online and offline?  The current inventory status is pretty unuasable.  (I've started using PRTG for critical devices)

I'm about 10% converted and can see it increasingly more difficult to manage once I have 300 or 400 devices online.

Are there other tools out there or how do you use it effectivly today?

Thanks,

Ryo


We have a home built CRM that we use to manage IP assignments and radio inventory. As far as monitoring, we have CNS in place, but use it primarily for syslog and firmware upgrades for the equipment. All of our customers are added to a cacti box to record information such as RSSI, SnR, etc. as well as traffic usage.

Alerting is handled by a What'sUp Gold server and we have all of the infrastructure added (APs, routers, BHs, power monitoring via SiteMonitors), as well as a selection of customers (4-8) from each tower so we know if there is a wide-spread issue. The server sends a text via email if anything stops responding within minutes so we can look at and address the issue.

We have also started using The Dude to get a bird's eye view of our backhaul traffic and tower routers. Each link we put up is added as a dynamic line that reports bandwidth at the current point in time, as well as keeping a record of the traffic so we can look back historically to see where bottle-necks may be occuring.


@Ryo Haines wrote:

It sounds like some of you have quite a variety of equipment in your environment.

I'm currently utilizing 2 vendors, UBNT and EPMP, and also looking at Mimosa (they look pretty impressive)

How has your experience been with varying hardware and do they seem pretty comparable?

It's tough for me to go and buy 2-3 different backhauls to test a link and see which one works best.

I settled on the force 110's for my backhauls and some airfibers when I need more bandwidth.

How do you go about testing making sure you get the best out of every link?

A lot of questions I know, but thanks for your feedback,

Ryo


I'd like to second what the other contributors have stated. We have been placing up ePMP 5 Ghz for new and replacement backhauls under 100Mb/s. We purchased a network running Ubnt backhauls and have seen consistent speeds of around 50Mb/s out of them using 20Mhz channels. The ePMP in our experience has delivered just under 70-80Mb/s at the same channel width.

Anything faster, and we plan according to the distance, speed required, spectrum available. We are using AirFiber links at a few sites, but have been primarily purchasing licensed links for higher capacity shots. 


@uberdome wrote:

This is more of a business question:

When transitioning your network to ePMP, how do you justify the cost of the new equipment and associated installations?

Which methods have you found to work to recover the costs (i.e., greater revenues due to selling higher speeds to customers or asking for an "upgrade" fee while keeping the same monthly rates)?

Thanks, Chris


In our case, we generally add a layer of ePMP or PMP450 to increase our offerings in that area. Most of our customers are asking for faster speeds and these lines are capable of delivering them while continuing to provide the features that made Cambium attractive to us in the first place like GPS and frequency re-use as described by the other contributors.

Our installations are contractual as far as service retention to help cover the base cost of the equipment without charging the customer full price up front. We allow them to change speed without renewing, paying fees in most cases. The higher throughputs allow us to sell higher packages recovering that cost quicker. We also have customers pay a one time $49 upgrade fee if their requested speed upgrade involves moving to the newer equipment. We use the same explination that Eric mentioned, the radios cost us so much, plus sending the technician onsite to install it, etc. Most customer's don't think twice, and then they start paying the new faster, more premium price (usually only $10-$15 more per package) to help us recover cost.

Our panelists will be re-joining the discussion in a few minutes. Feel free to post your questions and comments now.

I guess I have a questions for the other panelists and WISP's watching...

- What kind of traffic shaping do you use?

- Do you do any DPI (deep packet inspection) on your network?

- What are your thoughts on the new Title II regulations?
- What solutions do you use for CALEA, and have you ever had to procure data for a law enforcement agency?

- How do you deal with copyright notices, subpoenas, and users abusing bandwidth/torrenting?

Traffic shaping we use SM level QoS profiles for most our network.  We don't currently do any DPI, our main source of monitoring traffic is our Tik's at the sites, using Torch.  We have responded to requests / subpoenas for law enforcement.  Usually it's asking for a name associated with an IP, and we are able to comply.

What bandwidth ratios (sold:delivered) do you actually see on your networks?

How much does it vary for different service plans (I assume the ratio goes up with faster speeds, but how much)?

What ratio do you use for designs for new deployments?

Thanks, Chris

1 Like

That's a good question.  I know a lot of companies deal in 'burst' speeds for their marketing.  We have always tried to be honest with our customers.  So if we sell a 6MB service plan, we will allow a burst on that conneciton, but it will sustain 6MB.  That applies to all our service plans.  We build the network based on estimated clients, average speed plan, and technology capabilities.  If you overload a network, you build it bigger, that means you're growing.


@Eric Ozrelic wrote:

I guess I have a questions for the other panelists and WISP's watching...

- What kind of traffic shaping do you use?

- Do you do any DPI (deep packet inspection) on your network?

- What are your thoughts on the new Title II regulations?
- What solutions do you use for CALEA, and have you ever had to procure data for a law enforcement agency?

- How do you deal with copyright notices, subpoenas, and users abusing bandwidth/torrenting?


we traffic shape at the CPE,     we offer 3x1 , 6x1.5  9x2 and 15x3    ofter all of the epmp upgrades are done, we are going to double the speed offerings for no extra.

DPI, we do not currently but we've talked about it.

Title II, I'm mixed. I like the idea of a level playing field but i can't help but feel like its step 1 for more FCC control.  I haven't completely read the entire article yet, so my opinion isn't fully formed. 

Calea, we use microsoft 2008 DHCP clustering service, it maintains a log off what mac was assigned what IP and retains those logs for 1 year for us.

unfortunately yes, the DHCP logs came in handy at that point dealing with a copy right notice.  we've heard both ways about the legalities of blocking torrents ect. we've chosen not to directly stop based on the advice of our lawyer.

It's really difficult to just pull a standard over-subscription ratio out of a hat and stick with it. Through experience we've found what over-subscription works on what platform, but ultimately, you'll need to keep tabs on what throughput the AP is seeing and what the frame utilization is, is more important then a fixed number.

On PMP100, this was pretty easy, because an advantage AP would support around 14mbps net aggregate throughput (both up and down), and then based on your downlink % gives you the max theoretical download. So if your subscribers are connected at 2X modulation and have near perfect frame %'s... you could then look at your ethernet port statistics and see it plateau's at around 10-11mbps (typically during peak hours) also latency will increase across the board, so another indicator that you're reaching 100% frame utilization/capacity on that sector.

We really don't like burst plans either.


@uberdome wrote:

What bandwidth ratios (sold:delivered) do you actually see on your networks?

How much does it vary for different service plans (I assume the ratio goes up with faster speeds, but how much)?

What ratio do you use for designs for new deployments?

Thanks, Chris


we watch to see what the hardware is doing, we've got areas we have 50 people not touching an AP, and that same AP is being ran to its limit with just 25 in another so there isn't always a good 1 solution for over subscription.  we check in with the gear during peak times and look at usage histories from time to time to make sure its not getting blown up at peak.  

our Old PMP 100 gear is struggling to maintain old speed packages.  the 320 gear is keeping up with our old packages and generally on all of our upgrade areas, customers get the offered speeds 24x7 BUT we've also started to over build to look to the future IE  4 5g and 4 2g APs per tower, checking in on frame usage over night and adding as needed.      the UBNT gear is falling victim to self interference in areas and isn't holding up as well, which is why we are replacing with EPMP

I agree with the other panelists, to operate efficiently you really have to keep a theoretical finger on the networks pulse when it comes to speed plans, and not oversubscribing.  I will say though, always be conservative.  Buy yourself time to improve your network by not overloading it too quickly.  Start with lower speed plans, and add faster plans as you get a feel for the networks capabilities.

We've been able to squeeze quite a bit more life out of our PMP100 gear by doing some pretty advanced QoS and traffic grooming using some very specialized equipment. We basically pack all the lower end accounts on the PMP100 gear, and anyone that wants a higher level account, we work a deal to swap to ePMP or PMP450.

That's exactly how we operate also.  Our SOP is only 1MB and 2MB on 900mhz FSK, 1MB and 3MB on 2.4Ghz FSK, and then faster plans like a 6MB or 10MB plan is available on PMP320, PMP450, and ePMP.