Hey everyone! Author here. Happy to answer questions. Thanks for posting @tomcam!
Just a quick note about me - I currently work with the great folks at common.net running an ISP that is similar but not exactly like what's described here. Look us up if you're in East Bay maybe we can get you better Internet service! This site is not associated with common.net, though.
Hi, I've been doing this for ...um 15 years or so. Generally tried to exit the business because in our area it turns out to be a money pit. However, I notice you don't mention anything about interference : we found that even in a deep rural area this is a major headache. It can be hard to find even 40MHz of clean adjacent channels so delivering tens of megabits via P2MP is not so easy and that's before you look at the issues with air access control and half-duplex. The interference picture is always changing too so just because your network works today, doesn't mean it will tomorrow. You're made of stronger stuff than me to try it in an urban area!
Yes interference and channel planning can be a big headache. I do have planned a section that will cover this but haven't finished it yet. I've typically gotten around interference problems by 1) using licensed (or at least not 5ghz, preferably full duplex) backhauls and 2) using AP antennas with narrow beamwidths (higher gain, less area covered.) I'm sure you're doing the same things if you've been doing it that long.
Actually I didn't see anything about licensed spectrum on the site, but of course that helps with backhaul and high-end subscribers. For many years we really didn't see any interference issues with P2P links @5GHz but that has changed recently. We have been deploying dishes with shrouds to cut down on sidelobe gain.
My experience has been that interference is as much a problem at the CPE side (and much harder to diagnose there) than at the AP. I've also noticed that with Ubiquity/Mikrotik-grade radios in the 5GHz spectrum you can easily get interference from your own co-located radios at the same site, using different channels (because they are not so good about suppressing out-of-band radiation).
Shrouds should help against out-of-band radiation. I'm no expert but I assume that they do more harm in a single radio environment than without but they should help when you have multiple radios next to each other.
Not only out-of-band, but also in-band interference from local and remote transmitters (whether direct by transmitters operating on the same channel, or indirect via out-of-band harmonics transmitted by low-cost radios).
Hi Graham, nice site, love you are giving back! I've also done a few Ubiquity design/install projects.
Couple of things I'd suggest. (take/or leave) You are clearly super knowledgeable in all the products you've mentioned, but it seems like the audience for this site is the newbie trying to break into the industry. With that in mind, I'd try to stick to a single product recommendation, and deep dive into it. For instance, in the pick a hardware platform you've listed four options, consider picking one, and if you feel the need to dump your extensive knowledge of the others, then perhaps do it in another section/addendum.
Consider hosting this where you can build a community. Github pages does this sort of thing for free.
Good advice, thanks. I have struggled throughout with knowing when to be more specific and when to be more general, although I do try to pepper in a lot of 'if you don't know what to do, do this' type advice.
I've been trying to build the community around this matrix chatroom: https://riot.im/app/#/room/#startyourownisp:matrix.org I'm not really a developer by trade so Github isn't really a go-to solution for me, good thought though!
I disagree a little. I'd rather you overshare by default so we don't miss an opportunity to learn something. The kind of people who will actually go deploy ISP's vs talk about them will not fret over 4 vs 1 device. They'll just email you for clarification. :)
I do like the idea of simple, base recommendations with context-driven extras. Plus, in-depth guides on what you found each piece of equipment useful for, common situations you ran into with equipment or deployments, how you dealt with them. This kind of thing could be like a blog you do regularly as you go so it's not too time consuming. A small community could help you curate it or answer common questions on a forum or something.
Anyway, thanks for your efforts as I found all that info useful in putting together cost ranges for some people. Personally, I'm wanting something similar for fiber maybe like Sonic does or those folks in Britain rolling it out in rural areas. The wireless I'm collecting for others or just as a fallback option for any project I get into if fiber is too infeasible.
Thanks for the reply. Github pages used to be aimed purely at developers, but I think it's pretty mainstream now. If you've managed to get a web site up, then you won't have any problem with this. Happy to help if you have any questions PM me.
I checked out riot.im but honestly I found the interface really confusing. Also, the lack of integration to a major auth service like Google/Twitter/Github/etc is a real turn off for me.
> I checked out riot.im but honestly I found the interface really confusing. Also, the lack of integration to a major auth service like Google/Twitter/Github/etc is a real turn off for me.
Matrix is decentralized, using centralized auth wouldn't work.
You could even host your own matrix homeserver to talk in that room. (like email, or XMPP)
Gotcha, makes sense. I'm hosting on Firebase right now which seems fine and is close enough to free. Would there be a tangible benefit to moving to Github Pages? It wouldn't be difficult, it's a static site built with Hugo.
The main tangible benefit is that others can directly interact with each other and the content in one place using the built in collaboration tools.
It would be easy to move. Just upload the web site files to a Github repository with index.html in the root path, and turn on Github pages in the setting tab.
Other benefits that immediately come to mind are:
- Anyone can submit suggestions, these are called Pull Requests.
- Eventually you can add trusted community members as collaborators and they can edit pages directly.
- There is already a massive community that know this interface.
- It's free for public sites.
- Last but not least, it's super easy to edit the site, just do it directly in the browser, no need for builds, deploy scripts, etc.
Also there's this mailing-list (in english) dedicated to sharing knowledge between DIY ISP, a database and map of DIY ISP[2], this series of blog posts (in french) documenting how to build an ISP[3] and a wiki page centralizing resources about building a non-profit ISP[4] on the FFDN[5] wiki (Non profit ISP Federation from France).
Great info thank you! And thanks for the link back. All great info and I've had a lot of questions from people outside the US which is unfortunately where most of my experience is so these will be great resources to hand out.
Agreed that this is a better solution but the point of the site is to get from 0 to 1 customer as simply as possible and setting up a DNS server adds a complication. Another HN user recommended opennic, though - I might change the recommendation from Google to something like that.
I see. OpenDNS would be another, but I'm not sure who owns them (iirc they were bought by someone that made me go "ah, no more OpenDNS for me" but I might misremember). Opennic I didn't know of, I'll check that out next time I need a resolver somewhere :)
Hey Graham. Do you pay Comcast/AT&T to connect to their ISPs? If so how much? What is the benefit of setting up your own ISP if you need to connect to one of these giants anyways.
In some cases yes you are buying an upstream connection from Comcast/AT&T/Centurylink/etc. I actually have a note about that specifically on the site, as it happens! (https://startyourownisp.com/posts/fiber-provider/#overview second paragraph)
Things to note - you're not buying a regular residential or even a business line, you're (usually) buying a dedicated Internet Access line and dealing with an entirely different group for sales/support. It does seem kind of strange at first but I think the reality is that the raw bandwidth isn't one of the main costs for those companies, it's the support/last mile build&maintenance etc that's expensive to them - so if you're willing to give them some money and take that part off their hands they don't seem to mind.
Also - there are plenty of other companies selling fiber that don't also sell residential Internet products, so you're not always buying from them.
You'll need to do a lot of traffic analysis to make a cost/benefit analysis of purchasing international capacity.
Transit is likely to be cheaper until you get big enough to hire someone to do this analysis for you. At that point, they will probably look at purchasing some small amount of transit from a larger ISP (e.g. Megaport, Verizon, etc) to allow you to peer freely with your largest sources of traffic. E.g. if you have a lot of Korean families who stream a lot from Korea, it might make sense to get a gig or two of bandwidth to Korea, and putting a router there to peer with Naver, etc. rather than paying for those 1 - 2 Gbps of transit. But it's probably not really going to make sense until you're in the 10's of Gbps.
Good questions. I know very little about international fiber markets, unfortunately. I do have a fair amount of experience deploying LTE in fixed wireless environments, though. I'll work on adding some of that to the site and feel free to reach out to me with questions.
> what are the biggest complaints from your customers?
The other questions I think were answered below but this is a great one so I'll answer it directly. Biggest complaints that I currently see / would expect people to see:
1) No public IP addresses (this is addressed on the site and elsewhere in the thread)
2) No TV package, have to cut the cord
3) Can't provide service everywhere. "My neighbor can get it why can't I?" Well .. your house is 1 story buried in trees, unfortunately.
4) Sometimes wireless is just weird and there are things that don't work, especially when you're optimizing for a cheap home install. Some customers have a bad experience and as hard as we try we can't ever get it much better.
These numbers are from an 802.11ac based network using proprietary MAC (Mikrotik NStreme), YMMV.
Here's a 2 mile link @780Mbps modulation rate, tested with ping -s 1500:
72 packets transmitted, 72 received, 0% packet loss, time 71108ms
rtt min/avg/max/mdev = 1.051/3.132/26.567/4.466 ms
Here's a 10 mile link @650Mbps modulation:
122 packets transmitted, 122 received, 0% packet loss, time 121164ms
rtt min/avg/max/mdev = 0.984/2.174/20.479/3.260 ms
And finally here's stats over the two links combined (the path between my office and civilization):
68 packets transmitted, 68 received, 0% packet loss, time 67102ms
rtt min/avg/max/mdev = 1.755/4.471/14.968/3.658 ms
Note that if you allow too much traffic on these links they will saturate and QoS will go to crap. So careful traffic shaping is in order. They are half-duplex so achievable throughput is much lower than the modulation rate (these links are delivering me around 250Mbits/s).
At unlicensed frequencies (2-6GHz) snow and weather are generally not a problem. You do have to choose antenna design wisely : the satellite dish type antennas that have a separate reflector can get snowed up, but people here in snow country are aware of that issue and brush the snow away. Rain fade is not a practical problem : you need some link received power margin for reliable operation and the worst that happens is rain eats into that a bit. Lightning is a much bigger issue in my experience: you need to be aware of proper grounding techniques (which can be expensive) and you need to be prepared to replace some gear every year due to EMP damage. Also fyi you want to at all costs avoid deploying CPE antennas on a roof. The roof is a nightmare for access, insurance costs, risk of causing leaks, and potential injury. For that reason WISPs will try to mount on a eave, soffit etc rather on the roof surface.
My son uses our network for gaming and while he does complain quite often about QoS, when I investigate the problem is _always_ with our upstream provider or the server end and never to do with our network.
OP stated above that he is using licensed spectrum for backhaul (probably 11 or 18GHz) and therefore will have better latency (full-duplex radios and hence lower MAC overhead).
Hey! Just wanted to say that although you haven't come to my house yet, everyone I've talked to in Alameda loves you guys. Having worked at an ISP startup, I know how tough it is to succeed -- hope you guys manage!
I know this varies greatly based on location but it would be good if you could include a section on legal challenges given that there are many municipalities actively trying to discourage any sort of competition.
Secondly, looking at common.net, how did you arrive at 75Mbps as the offering speed?
I actually have in my plan to include a section like that, not quite there yet though. But in general legal challenges are less of a problem than you might think - most of the legal challenges you hear about with running an ISP come from things like who owns/is allowed to use conduit, right of way for trenching new lines, those kinds of things. With wireless you sidestep all of that.
75mbps is what we feel like we can reasonably achieve across our entire user base. Most of our customers are up to double that.
Also I should add - what I've found is that the municipalities are somewhere between neutral and really excited to have another provider in town, as long as you're not putting too much ugly equipment on people's houses. Sometimes the municipality is a really great partner and will help find relay sites, etc. It's the incumbent cable cos who don't like the competition.
Hey Graham, I'm in Oakland in a six story building downtown. Currently have DSL with Sonic, but am getting impatient with their fiber rollout in the East Bay. Any chance of getting service from Common outside of Alameda?
Yes. Sign up for the service on the site, that will get you in the queue. Also email me with your address (graham at common.net) and I'll take a look at the building specifically.
Wow, I've been shopping for WISPs in the South Bay and pricing seems to be an order of magnitude higher ($20/Mbps/month or higher). What your secret? Shorter links and higher density?
How dependent is this on flat land and no trees? This would be great for northwest Arkansas, but we have lots of trees and hills. We are stuck with only Cox, and ATT in some places.
Generally you need unobstructed line of sight, so you need to go around or above trees. Sometimes you can mount an antenna _on_ a tree, but only sometimes.
Unfortunately, to rain on the parade a little: note that it is hard using wireless (due to technical limitations to do with Shannon capacity) to compete with a copper coax provider. It really doesn't work notwithstanding the OP's claim that they can deliver 75Mbits/s to subs. I really don't know how he is doing that unless they are all fed with dedicated point-to-point links (which won't scale and costs $$).
Not surprising, although I still believe their could be opportunity in the rural areas, whose only option is satellite, which is both expensive and slow. Just getting that line of site is difficult. Thanks for the info!
I noticed the part on IP, is it true that the termination points for a customer do not have a public ip address? Isn't that an issue for your customers?
It is an issue, and it's a tricky one to solve. It's getting much harder to get a hold of public IPv4 addresses and lots of smaller providers are using NAT instead. It's not a total show stopper - the vast majority of customers don't even notice and only a few actually need it enough to not use the service, and for those you can probably come up with clever ways around it (like a VPN) but it's definitely a problem. I think the best way out is IPv6 (but of course we've been saying that for decades).
Just be ready with a story when Law Enforcement show up asking you to identify whoever used one of your IPs to commit a crime on such and such a date in the past.
That's true, I've been there! I've never had them be unreasonable about the situation but maybe I've just been lucky.
The real scary one was servicing a relay site on top of a bank late at night and the cops showing up guns blazing thinking the place was being robbed ;)
I had US Secret Service show up at my office once, demanding information on an external NAT IP. We had 10k NAT'd devices behind that IP. Apparently some doofus wrote on some govt. website they were threatening the president. A warrant later (which took < 1hr to get!) I told them everything I could, which was "Oh that traffic was to this entirely different legal entity, you need to go talk to them now."
Good question. Off the top of my head: 1) much better wireless performance, higher modulation at lower receive signal levels. 2) easier channel planning - you use the same channels across all your towers and the core figures it all out. 3) you can potentially get roaming agreements with carriers (I don't know of anyone who has successfully done this but apparently it's feasible)
Big disadvantage is cost - you have to run a core which is very expensive ($100k+) Or you can use BaiCells and they'll rent you use of their core for fairly cheap.
I would very much recommend against this - off-the-shelf core networks are shit. Look up "SS7 vulnerabilities" on Google to see what I mean... and yeah your LTE core network would be based on that.
LTE itself is good but you'd need to build a proper core network for it - at the moment none of the off-the-shelf gear you can buy is any good.
What an amazing thing you've done taking the time to share your knowledge for free with everyone - thank you!
This is very timely for me. I actually had just started paying a consultant to help me plan out a WISP, because it's so overwhelming starting out because you don't even know what questions to ask to get started (the internet is a great resource for so many things... but much less helpful when you don't have well-formed questions!). So far everything I'm seeing lines up with what I've researched myself or what my consultant has helped me with, so this seems to be an amazing free resource!
I even already learned something in the first couple sections - I had been using Google Earth Pro to scout potential tower locations, and I knew what a viewshed was, but I thought I was going to have to pay for ArcGIS or another program to generate them. For some reason I thought Google Earth Pro could only do elevation profiles! It's cool you mentioned using a drone; I was already thinking of doing that and it's awesome to know that's not a weird thing to do. Have you heard of drone photogrammetry[1] before? It's probably overkill but it seems like an interesting concept for generating high resolution viewsheds (i.e. taking foliage and other obstacles that won't show up on a topographic map into account).
My consultant helped me find a fiber backhaul solution; CenturyLink fiber dominates my area so I'm either going to be buying transit directly from them or buying transport to an internet exchange at the nearest metro area (the latter sounds way more fun to me and has some interesting peering options). But I'm still at the point of evaluating the customer base. So one thing that I'd like to see is more details on how to make first contact with potential customers to gauge their feelings towards a new ISP. So far I'm thinking of buying a list of mailing address from a mailing list broker, then mailing out postcards with a link to a Google Forms survey (and maybe an alternate option of just calling me if they don't currently have internet service). But I don't know how this is typically done.
I'd also like to see a section on skillsets that an ISP operator should develop. I started studying for the CCNA because I want to be able to wrap my head around the carrier-grade networking that will belie the business. It would be interesting to hear your opinion on how much you can punt on these kinds of skills vs. what you really need to know, or even when you should just hire the work out.
I also signed up to take a course on fiber optic outside plant planning because some day I'd like to level up to being a FTTH ISP. That's a whole different ballgame since you basically have to operate a construction company to plow/blow/splice all the fiber, deal with all the permitting, insane CAPEX etc. I know that's getting way ahead of myself and I should just focus on only WISP-related things to start but that's what I wish I could do if I were a millionaire. But if you have any experience with fiber ISPs I would be really interested in seeing that on your website as well (later!).
Another cool topic to cover (sorry if I missed it) would be the whole process of putting up an actual tower - I've heard that it can often be as cost-effective to erect your own rather than leasing (like pays for itself in 2-3 years). I was just reading a series of blog articles[2] a couple weeks ago by a guy who put up a ham radio tower that had lots of very interesting info and pics.
There is free/open source software that will compute viewsheds for you, and data downloadable from the USGS. In times past I did a lot of this but really you just need to go look with a telescope or a camera on the end of a long pole. But note that the work you do to figure out if you can provide service to a prospective customer can easily exceed the lifetime value of that customer!
The Post Office (in the US anyway) will mail a postcard to every household in a defined area for reasonable cost.
Remember to get good insurance if you construct or climb towers.
Honestly I really question your sanity though. Think of all the people who decide they are going to open a restaurant: we shake our heads at the large subset of them who have absolutely no restaurant experience. We know that in order to make money at a restaurant you at least need to absolutely nail the whole making food that people like to eat and making the place pleasant and inviting and all that (execution). If you don't have that, you're doomed for sure. And then some large proportion of even the experienced restaurant folks still fail -- we look at their efforts and shake our heads "well they should have known this town doesn't have a big enough market for a fine dining place", and so on.
As someone who does know all about how to build and run a wireless ISP, I'm reading your post like how most folks look at a doomed new restaurant opening.
Sorry to be a party pooper but I figure I may save you some $$$. I think at least you need to be doing this in a place where there is _no_ competition besides satellite. If there is even just regular LTE service people will scrape by with that rather than pay you for better service. If there are traditional providers (Coax/HFC, DSL) then forget about it.
Just a quick note about me - I currently work with the great folks at common.net running an ISP that is similar but not exactly like what's described here. Look us up if you're in East Bay maybe we can get you better Internet service! This site is not associated with common.net, though.