Let's discuss "being your own ISP."

[Dimi3os]

DISCLAIMER: this is all hypothetically speaking!

 

Hey guys, after doing some research the past days about what you would need to be your own ISP (as a matter of fact people do this in rural areas regularly).

 

Well, since I live in an area that's quite urban there wouldn't really be a need. Unless you factor in the relatively slow internet speeds, the lack of fiber (in Belgium at least) and the bad service.

 

It would be possible to set up a wireless (mesh) network but since I don't live that long from a data center (let's say 1.5 miles) and the data center is located pretty much near the center of the city where I live in (~100 000 inhabitants), using micro-trenches it wouldn't be too difficult to supply the whole city center and the "banlieu" (= the edge) with cables/WiFi access.

 

If there would be a company (that isn't about milking out it's customers) to provide this together with a more personal approach would you go with it.

 

Sadly, it might be hard to get the big ISPs to allow this, luckily Belgium isn't one of these countries that isn't open for inovation so I guess this idea could be feasible.

 

 

[Sumsar]

Create your own isp company:

 

step one: win 2 million €

step two: buy equipment

step three: enjoy your internet speeds

 

 

 

 

Cool idea though. Try and contact your isp with the suggestion.

[Dimi3os]

Create your own isp company:

 

step one: win 2 million €

step two: buy equipment

step three: enjoy your internet speeds

 

 

 

 

Cool idea though. Try and contact your isp with the suggestion.

Latong fiber isn't that expensive and hirong your own bandwith from a data center that connects to a higher tier provider would be great, especially with the recent net neutrality debate that's starting to pop up in Europe. You could keep the internet neutral

[Altecice]

Not quite as simple as that, your ISP will need to have more than 2millions worth of stuff to work well if it works at all... Without your own DNS servers and those being registered to the root DNS servers on the internet all you will have is one massive (fast) LAN.

[lubblig]

Not quite as simple as that, your ISP will need to have more than 2millions worth of stuff to work well if it works at all... Without your own DNS servers and those being registered to the root DNS servers on the internet all you will have is one massive (fast) LAN.

I have no idea of how ISPs work but if you ignore the legal part of it, could you not just use something like opendns.com to be the DNS servers? Since I on my own can change DNS from the one my ISP uses in my router to something else, wouldn't it be possible to just use something publicly available? (But again, ignoring that you would have to go through legal stuff to, as an ISP, use someone else's DNS servers.)

 

Again, I'm just speculating with the basic knowledge I have... I really don't know much about how ISPs work and what kind of equipment is used... Everything was hypothetical anyways so

[Altecice]

I have no idea of how ISPs work but if you ignore the legal part of it, could you not just use something like opendns.com to be the DNS servers? Since I on my own can change DNS from the one my ISP uses in my router to something else, wouldn't it be possible to just use something publicly available? (But again, ignoring that you would have to go through legal stuff to, as an ISP, use someone else's DNS servers.)

 

Again, I'm just speculating with the basic knowledge I have... I really don't know much about how ISPs work and what kind of equipment is used... Everything was hypothetical anyways so

 

You would still be required to register your DNS server(s) with the main TLD DNS servers so that other DNS servers can contact it when a request is sent from something outside of your ISP to one of your users.

[Dimi3os]

Not quite as simple as that, your ISP will need to have more than 2millions worth of stuff to work well if it works at all... Without your own DNS servers and those being registered to the root DNS servers on the internet all you will have is one massive (fast) LAN.

Not quite as simple as that, your ISP will need to have more than 2millions worth of stuff to work well if it works at all... Without your own DNS servers and those being registered to the root DNS servers on the internet all you will have is one massive (fast) LAN.

Let' say that 100 people would subscribe to the system, if you wojld hire space and bandwith from a data center and set up a server there that would go over these 100 people and would manage the "LAN" network, wouldn't this be a lot more realistic. As suggested in m original post, people can do this on rooftops in NYC.

[Altecice]

Let' say that 100 people would subscribe to the system, if you wojld hire space and bandwith from a data center and set up a server there that would go over these 100 people and would manage the "LAN" network, wouldn't this be a lot more realistic. As suggested in m original post, people can do this on rooftops in NYC.

 

That would work, depending on the service you buy however it would still only be an internal-LAN. This means you would run into other issues such as Broadcast packet storms. You would also run into licencing issues on the legal side.

[dzonidev]

You would need a big investment either way. If you go Wireless like 5Ghz you will hit a big limit at ~100Mbps, sure the cost would be much lower and *easier* to setup. But you would need to deploy a lot of access points because one realistically could only support like 15 clients at those speeds.

Which returns you to the fiber optic problem with micro trenches, licences are not cheap, and running fiber is expensive. Because you will need to terminate it, setup junction boxes, etc. And not to mention the part when you need to install it into a clients house.

 

These are all issues that can be easily solved with a good investment.

 

As for the gear, you will need at least 300k for equipment, routers, switches those GPON things, etc... Not to mention your options are limited to Cisco which costs an arm and a leg or some chinese supplier like Huawei or DCN.

 

And your biggest obstacle bandwidth. If you want to break even in the first year you will need a lot of bandwidth,  a 10 Gbps backbone can essentially feed 130 (100Mbps) clients (yes over provisioning) which guarantees ~80Mbps for every client with the ability to go up to 100Mbps. In my opinion this is a hell of a service.

And even then you can easily squeeze in more clients because nobody uses 100% all the time. Even peak times wouldn't reach 99%. Which leads me to the next part and that is the cost of the fore mentioned bandwidth, a big carrier will usually bill you around 1$/Mbps (optimally) which comes around 10k/Month for a 10Gbps port. And if you bill your clients $50 for a 100Mbps connection, you earn 6.5k.  So you can't even pay off the bandwidth, unless you over provision which a lot of ISPs do, thus providing a bad service.

 

How do ISPs make it then? Well they buy 50 Gbps get a price of $0.4/Mbps and they almost double the profit.

 

I could go on and on, about this topic. So any questions?

[Dimi3os]

You would need a big investment either way. If you go Wireless like 5Ghz you will hit a big limit at ~100Mbps, sure the cost would be much lower and *easier* to setup. But you would need to deploy a lot of access points because one realistically could only support like 15 clients at those speeds.

Which returns you to the fiber optic problem with micro trenches, licences are not cheap, and running fiber is expensive. Because you will need to terminate it, setup junction boxes, etc. And not to mention the part when you need to install it into a clients house.

 

These are all issues that can be easily solved with a good investment.

 

As for the gear, you will need at least 300k for equipment, routers, switches those GPON things, etc... Not to mention your options are limited to Cisco which costs an arm and a leg or some chinese supplier like Huawei or DCN.

 

And your biggest obstacle bandwidth. If you want to break even in the first year you will need a lot of bandwidth,  a 10 Gbps backbone can essentially feed 130 (100Mbps) clients (yes over provisioning) which guarantees ~80Mbps for every client with the ability to go up to 100Mbps. In my opinion this is a hell of a service.

And even then you can easily squeeze in more clients because nobody uses 100% all the time. Even peak times wouldn't reach 99%. Which leads me to the next part and that is the cost of the fore mentioned bandwidth, a big carrier will usually bill you around 1$/Mbps (optimally) which comes around 10k/Month for a 10Gbps port. And if you bill your clients $50 for a 100Mbps connection, you earn 6.5k.  So you can't even pay off the bandwidth, unless you over provision which a lot of ISPs do, thus providing a bad service.

 

How do ISPs make it then? Well they buy 50 Gbps get a price of $0.4/Mbps and they almost double the profit.

 

I could go on and on, about this topic. So any questions?

What would you need to beat 240 mbps which is the best you can find for a small business in Belgium. It should be easier to do this with fiber right? At the moment isps run coaxial cables (with a fiber optic backbone (can't really be certain about this)).

[KishMe]

Have you looked into Ubiquiti gear? You could do this wirelessly (providing you have some line of sight to your clients or are mounting the radios above roof-tops) fairly inexpensively. Tons of people already do it, maybe check out the Ubiquiti forums. 

[dzonidev]

What would you need to beat 240 mbps which is the best you can find for a small business in Belgium. It should be easier to do this with fiber right? At the moment isps run coaxial cables (with a fiber optic backbone (can't really be certain about this)).

 

Okay let's do some math.

 

Say that the ISP you are comparing to bills $85 for 240 Mbps. That's $0.35/Mbps.

 

You could come to a price of $85 for 300Mbps. At that rate you will bill your clients $0.28/Mbps, which means either over provision or raise the price.

 

If you rent 5x10Gbps you get total of 50,000Mbps which is good for 170 clients without over provisioning and 200 clients with over provisioning. An ideal price you could come up with from a low level carrier like Cogent would be 0.3/Mbps, maybe even lower but you'd have to find a good account manager. With that price, you would be paying $15k/Month. Billing 200 clients for 300Mbps will earn you $17k. Which leaves you with $2k for maintenance and other things like the electricity bill, IP addresses, staff, etc.. And you still wouldn't break even.

 

But to get to this, you need to go through hell. Setting up 200 clients will take months. That's why now days in developed countries to be a successful ISP you need to start BIG. ​These are tiny numbers but goes just to show you the logic behind it. As someone said, a couple mill would get the ball rolling. An ISP here in Serbia invested $40,000,000 into the infrastructure. They offered FTTH only, but they quickly closed down because there were legal issues with the micro trenches, and monopoly ISPs made sure that the new ISP get's hit hard for this. They offered speeds like 24/12Mbps for as little as $14 USD, all fiber optic. You get a fiber gateway, everything.

[Dimi3os]

Okay let's do some math.

 

Say that the ISP you are comparing to bills $85 for 240 Mbps. That's $0.35/Mbps.

 

You could come to a price of $85 for 300Mbps. At that rate you will bill your clients $0.28/Mbps, which means either over provision or raise the price.

 

If you rent 5x10Gbps you get total of 50,000Mbps which is good for 170 clients without over provisioning and 200 clients with over provisioning. An ideal price you could come up with from a low level carrier like Cogent would be 0.3/Mbps, maybe even lower but you'd have to find a good account manager. With that price, you would be paying $15k/Month. Billing 200 clients for 300Mbps will earn you $17k. Which leaves you with $2k for maintenance and other things like the electricity bill, IP addresses, staff, etc.. And you still wouldn't break even.

 

But to get to this, you need to go through hell. Setting up 200 clients will take months. That's why now days in developed countries to be a successful ISP you need to start BIG. ​These are tiny numbers but goes just to show you the logic behind it. As someone said, a couple mill would get the ball rolling. An ISP here in Serbia invested $40,000,000 into the infrastructure. They offered FTTH only, but they quickly closed down because there were legal issues with the micro trenches, and monopoly ISPs made sure that the new ISP get's hit hard for this. They offered speeds like 24/12Mbps for as little as $14 USD, all fiber optic. You get a fiber gateway, everything.

I really appreciate your elaborate answers, it's so typical that the big ISPs do these things (we really need more companies like Tesla).

Have you any information on where you can study this topic, as in what you would need server wise, what each server and link in the chain would do etc. Even if you don't, I really appreciate your insight!

[Dimi3os]

Have you looked into Ubiquiti gear? You could do this wirelessly (providing you have some line of sight to your clients or are mounting the radios above roof-tops) fairly inexpensively. Tons of people already do it, maybe check out the Ubiquiti forums.

Thanks for your response, I'll make sure to check it out tomorrow after a good night rest