How ISPs work?

vivek.krishnan

BLR~ZRS-TX-1-MX
Skilled
I think most guys have not understood how the entire system works - whether its ISP offering internet speeds, or a mobile operator or airlines/hotel, or even your LAN network - its oversubscription. It's a simple concept - you look at typical vs peak usage.

If you are a ISP - you may have a high speed line of say 100 Mbps. With this, you can service around 100 connections at 1 Mbps each. However, on average, you will notice that most connections are utilized to the tune of 20-30% only. So, you have usually 70-80% capacity free on average. So, you oversubscribe and sell instead of 100 connections, say about 200 connections. This way, you double your sales without increasing your cost. Most ISPs now, do a combination of local peering + transparent squid caching + torrent caching to benefit - but these are usually Tier 3 ISPs like your cable operator.

Similarly, your mobile operator does not build towers with capacity to serve 100% calls, but the average. That's why during heavy usage like calamities, the networks fail.

The airlines/hotel operators do the same and assume people will cancel. They will pocket the termination charge as well.

Coming to a live scenario - your home or office local network, assume each device can pull 30 Mbps over WiFi and 100 Mbps over LAN, do you ensure your local server has 30x and 100x the speeds? It will peak at 100 Mbps/1Gbps/10Gbps (rare). If each device starts pulling data, then the network speeds go for a toss. Or look at WiFi - you don't get the most fastest WiFi AP even if you have several WiFi devices.

My point for mentioning this is that we have seen a spate of new operators who are providing 5/10/20/100 Mbps at cheap rates. Is it viable? Maybe. Is it viable if everyone starts downloading like crazy? Nopes. Thats why the terms will say upto X Mbps. And in many cases, FUPed.

Another point is about reduced upload speeds - ISPs offer lower upload vs download speeds on most plans for two main reasons - they keep the upload speeds for business users who might host things and to ensure that you can host anything with good speeds - because then you won't take a business plan. For most end users, they hardly upload stuff, except the occasional image/attachment/backup/etc, so it does not matter.

I have a feeling this should be made into a new thread in the ISP section - Maybe how they work? :p
 
Ok thats great and all but how much do I need to invest if I want to start providing BB internet connection to other consumers? Where and how do I get the main parent line? :p
 
Ok thats great and all but how much do I need to invest if I want to start providing BB internet connection to other consumers? Where and how do I get the main parent line? :p
Did you get any answer to your question?
I am also looking for information on setting up ISP and finding out all possible details.
If you come across anything relevant, please share.
 
Nope. Still waiting for @vivek.krishnan to reply.
And as for the ISP I believe you buy a dedicated line and bandwidth from another ISP and then redistribute it.

Uh, I am no ISP guy, but this is what I mentioned to @Ajay K Meher

If you want to setup an ISP, there are several various costs, some of which are fixed and some variable.

The main costs for you will be

1. Cost of bandwidth from main ISP
2. Cost of main router which will serve all your customers. This should also include redundancy.
3. Cost of connecting people from your main router till their homes. This would include wiring, switches and any extra equipment.
4. Cost of technical support - HR
5. Cost of software to maintain accounts and etc.
6. Upgradation costs - you may want to look at that
6. Cost of peering per GB.

At the drop of the hat, this is all I can think of.
 
Last edited:
Thanks vivek, that helped. Around how much $ do you think the whole thing will cost on an average?
Btw you were so enthusiastic you forgot to remove your phone number. :p
 
Thanks vivek, that helped. Around how much $ do you think the whole thing will cost on an average?
Btw you were so enthusiastic you forgot to remove your phone number. :p

Was getting late, so just copy pasted and forgot to remove the damn number.

Anyways, I would say a starting figure of about 10-12LL should help getting started, with wireless instead of fully wired. There was someone who mentioned they are running their ISP in south India, they could help. He runs it parallely to his fathers cable operator business.
 
I see. Thanks. It will be more in my location though. Rough mountainous terrain is also making things even more difficult. However if it can somehow be done wirelessly just like DTH (main connection to a huge satellite antenna and then wirelessly to consumers using receiverjust like how Tikona does it, i believe they call it WiMax) it can work. But it shouldn't have high latency. 3G is not an option. Will this be possible? BSNL also offer WiMax here but you know... >.>
 
If you want to start small, here is an idea on shoestring budget.

1. Buy bulk bandwidth from a service provider (say 50-100 Mbps).

2 Take it into a Linux box (with lots of network cards) which runs your applications (NAT/DHCP/accounting/PPPoE etc). Run it as a router.

3. Get a few VLAN aware switches which have fiber ports (for long haul). Chain them using fiber links. Give each customer a VLAN and extend the VLAN upto a port on the Linux box. Each customer now becomes an logical interface on the Linux machine.

A switch like this might work:
http://www.amazon.com/Netgear-GS724T-ProSAFE-24-port-Gigabit/dp/B00I5W5EGA/

4. You can implement FUP on Linux using simple scripts (poll the interface stats). Or you can use QoS to give unmetered bandwidth with CIR/PIR.

However if it can somehow be done wirelessly just like DTH (main connection to a huge satellite antenna and then

DTH (or any satellite links) work well for one way transmission. Sending data to a satellite is expensive (and the latency is high). I knew one person who used to sell a solution where the download comes through satellite and the uploads go through a phone.

BSNL also offer WiMax here but you know... >.>

See if you can get point-to-point WiMAX (or any other wireless solution) for connecting switches. Distribute locally using copper. Also remember that wireless deployment almost invariably need govt approval. Probably not good for the small guy.
 
Last edited:
Buy bulk bandwidth from a service provider (say 50-100 Mbps).
Some concerns there: what if the ISP breaks down? Won't my service be also affected?
Also I don't know how much bandwidth they can provide but lets assume they give a dedicated leased line of 100mbps connection. How many users will that be able to accomodate without altering the actual bandwidth from my side. Say, 1mbps/user so a 100 users in total or can it be able to accomodate more? Well more than 95% of the consumer base here might not know about the speed which they are assigned. :p




See if you can get point-to-point WiMAX (or any other wireless solution) for connecting switches.

How good or reliable will that be?
Rest of the points which you mentioned are all greek to me. :D
 
Some concerns there: what if the ISP breaks down? Won't my service be also affected?

Reliability costs money. What I had described was more of a guerrilla style ISP in an underserved area. If you are worried about ISP breaking down, connect to two ISPs and balance your load between them. Should be doable in linux (read on ISP redundancy).

How good or reliable will that be?

The technology is quite mature. Again, if you want protection from failure, add another link.

How many users will that be able to accomodate without altering the actual bandwidth from my side. Say, 1mbps/user so a 100 users in total or can it be able to accomodate more? Well more than 95% of the consumer base here might not know about the speed which they are assigned. :p

I would put 500 customers(or even more) on a 100 Mbps link. Most people will be happy with 256kbps/1Mbps (confirmed/peak). If someone wants more, they pay more.

Besides, business class links are symmetric (upload=download). Since most retail consumers are only interested in download speeds, the upload bandwidth can be sold separately (e.g for people interested in video calls, or some small business running a web server)

Rest of the points which you mentioned are all greek to me. :D

If you want to run an ISP, you should familiarize yourself with networking concepts. You could get a turnkey solution (after all, many cable TV providers also provide internet), but that will cost you more money.
 
Reliability costs money.
Yeah I understand. But two ISPs is not an option in my case though.
I would put 500 customers(or even more) on a 100 Mbps link. Most people will be happy with 256kbps/1Mbps (confirmed/peak).
Holy freaking shit. LOOL! And I was thinking 200 users at max haha. But its certainly doable. Most of the consumers here still rely on 64kbps 2G internet.
TIL a good conscience is bad for a good business. :p
Which option sounds less worse: advertise the plan as 256kbps and give 256kbps, OR, advertise it as 1mbps and give only 256kbps?
Since most retail consumers are only interested in download speeds, the upload bandwidth can be sold separately
True.
If you want to run an ISP, you should familiarize yourself with networking concepts.
Okay where do I start. I have some basic ideas about LAN.
 
Okay where do I start. I have some basic ideas about LAN.

I am afraid that is not going to be enough. If you are going to do it cheap as described by "dpandey" then you should be quite good with networking and linux, or hire somebody to do it for you.

I personally feel rather than the technology hurdles, it is the distribution that is going to be an issue. It would be nice (and cheap) if you can piggy-back on somebody's existing copper distribution lines, like local cable wala.

You can think about redundancy etc once you get this off the ground.
 
If you want to start small, here is an idea on shoestring budget.

1. Buy bulk bandwidth from a service provider (say 50-100 Mbps).

2 Take it into a Linux box (with lots of network cards) which runs your applications (NAT/DHCP/accounting/PPPoE etc). Run it as a router.

Use pfsense as along with captive portal/pppoe server for startes

3. Get a few VLAN aware switches which have fiber ports (for long haul). Chain them using fiber links. Give each customer a VLAN and extend the VLAN upto a port on the Linux box. Each customer now becomes an logical interface on the Linux machine.

A switch like this might work:
http://www.amazon.com/Netgear-GS724T-ProSAFE-24-port-Gigabit/dp/B00I5W5EGA/

I would advice against VLANs at the moment. Maybe yes when you are sure that extra investment is worth it. If you are getting managed switches, go for it.

4. You can implement FUP on Linux using simple scripts (poll the interface stats). Or you can use QoS to give unmetered bandwidth with CIR/PIR.

Yes, but would advise against implementing the same initially, but when you reach mass adoption, go for it.

DTH (or any satellite links) work well for one way transmission. Sending data to a satellite is expensive (and the latency is high). I knew one person who used to sell a solution where the download comes through satellite and the uploads go through a phone.


See if you can get point-to-point WiMAX (or any other wireless solution) for connecting switches. Distribute locally using copper. Also remember that wireless deployment almost invariably need govt approval. Probably not good for the small guy.

Have replied in quotes - these are just my views.[DOUBLEPOST=1457437618][/DOUBLEPOST]Redundancy - This is how most guys do it -

70% main ISP and 30% backup ISP. But that should not even be on your priority list initially, but after 6-12 months.

Ideally, 50-50 is best bet and you should load balance both. I do this with my ISPs from whom I take internet lines.
 
It would be nice (and cheap) if you can piggy-back on somebody's existing copper distribution lines, like local cable wala.

Can't. Local cable guy personally told me that they will offer internet services "soon." So the competition will be between them V/s BSNL V/s another private ISP V/s 2G/3G/4G network.

Just a quick reminder, my comments are more for general knowledge and information rather than the business part. Just because I ask a lot of questions please don't think that I'm ready to put it into action lol. Not really serious about starting an ISP business nor do I think that it will be feasible if the service isn't good=proper investment which i don't have.
But please keep the discussion going. Thanks.
 
Last edited:
I am afraid that is not going to be enough. If you are going to do it cheap as described by "dpandey" then you should be quite good with networking and linux, or hire somebody to do it for you.

I personally feel rather than the technology hurdles, it is the distribution that is going to be an issue. It would be nice (and cheap) if you can piggy-back on somebody's existing copper distribution lines, like local cable wala.

You can think about redundancy etc once you get this off the ground.

Actually you cannot piggyback on someone else's line if you are planning to offer ISP services.
 
I had posted this in the 'My Broadband' thread - I have come across a wireless service provider by the name of Airjaldi who are specifically targetting rural hilly areas where line-of-sight communication is not a problem. They buy their bandwidth from Airtel. Earlier they were a NGO, but they have started commercial distribution as well now. Their plans are really great, especially for people like me who have moved to rural areas where BSNL was the only option.
I pay 1100 for 2Mbps till 50GB and 1Mbps beyond that.
I will soon move to their 2700 plan which will give me 4Mbps till 100GB and 2Mbps beyond.
They charge 2-3 times that rate if you are a business owner. Most of their commercial clients right now are home-stay/motel/guest house owners.

They are now franchising out.
 
Last edited:
Back
Top