Bandwidth Cap Analysis: Why Bandwidth Caps Are Robbery

[mgcarley]

1.)Arent many of these are one time paid.. I mean like real-estate , equipments , laying of fibres..? also not to forget the deposits we pay , just in case and in many cases the modem is rented one , so it goes back to the ISP. Also we all know these modems to the ISP come at dirt cheap prices . that too on bulky orders ... they come even cheaper and a customer who uses it for >6 months would have got you back the cost of modem and will from there onwards reap you profits.


Oh dear. Welcome to Economics 101.

I'm not talking about equipment that the users ever see, but all of these things have ongoing costs. It depends on the ISP, but even if we buy/build real-estate there are ongoing costs for it such as maintenance, taxes or mortgage payments, and even if it is a one-time cost (that is, the company pays for it up front), that cost has to be amortized over some period of time. In the case of smaller ISPs, the real-estate is rented and thus has a fixed monthly cost.

Cables may have a one time cost but as with real-estate, that cost has to be amortized, and even then, the payments to the municipality are also recurring, and even then, the cost of the actual cabling is minimal compared to the cost of the payments to the authorities and the civil works (where the burying of cables is concerned, that is), but if I spend 1000-crore on my network today, I'm expecting to get that money back over the period of say 5 or 10 years, not 1 month.

You are right that the typical ADSL modems provided by most ISPs are dirt cheap and that at least half the cost is paid for by your deposit. This is where we in particular at a disadvantage: our modems are way not cheap, but we don't want to charge a Rs5000+ deposit!

2.) Lets be even too .. even if you sell a unFUPed plan of say 1mbps @ 2000 Rs or so , how many of them are going to use it fully ? Not to forget , i am on a 512kbps plan and i am supposed to get download speed around 64kbps , but i get an average of 50~55kbps , so already the connection is degraded . its usually said maximum of 80% of the speed is sure confirmed ,a customer can get on a connection. virtually on 1mbps plan , when its supposed to be 128kbps , he will be getting around 90~100 kbps. So full quota can never be used.

The same argument can be put forth that if the FUP is calculated correctly, most users should never reach it. As for the degraded connection, this is one of those things that you must put up with in the world of the copper (or wireless) last-mile.

3.) All i am saying is instead of advertising it as a 4mbps plan with FUP , why cant ISP simply give a straight forward plan with xxxx Rs ? This gives customers like us , a big relief while browsing , instead of worrying when we will reach the cap. Indian net users are willing to pay preimium for good speed with no limits. I paid 1500 RS + tax for 512kbps UL 1 year back and to be honest , AFAIK , i was the one customer in my village to have subscribed for it.


I don't disagree with you there - the way the plans are advertised in India has often felt wrong to me in the past because the FUPs were so tiny. However, with caps slowly becoming somewhat more "reasonable", this is becoming less true and again, if they are calculated correctly, then the majority of users should not reach it and in such a case only the heaviest of downloaders should be affected.

Also, the fact that you were the only one who subscribed to the Rs1,500/month plan says something too: the ISP is going to have to get a lot more subscribers paying for the Rs500 plans to break even. From there, the numbers are a juggling act as to how much is needed to supply the whole village etc.

In our case, we need something like 250 subscribers to break even on fibre in a given area, and that's at our prices: forget the sub Rs1k prices other ISPs have. Fortunately, in metros this is quite easy because of the density, and in villages it's not as bad because the costs to the municipality (etc) are less, so a larger geographical area can be covered for the same amount of money (which ends up balancing out the numbers).

4.)as for the second quote , i missed the the word "bussiness plan"


Fair enough.

5.) Lastly , If the pipes are owned by FUPtel and reliance , if my memory serves me right , about 80% is owned by FUPtel. If that is the case , they have all other ISP by thier throat and all other ISP have to dance to FUPtels tune. ITs not like govt is distributing the bandwidth. If thats the case , we can argue the prices are high . Why should we believe that price for 1mbit/s is Rs2,200 ? when the price is determined by a private company ? FUPtel can easily CARTEL with others to keep prices high to reap profit .. cant that be true ?


I think you didn't read what I wrote: I wrote that Rs2,200 might be some fictional cost of *delivered* bandwidth - that is, including all the equipment, staff, last mile etc. In real life the numbers are different but of course they also vary widely based on how well the ISP has negotiated with the upstream provider and how much they buy. The cost of *wholesale* bandwidth is under Rs1,000/mbit/month for the big ISPs but that only gets it as far as their offices in either Mumbai or Chennai (and it costs more for smaller ISPs). The distribution network that takes it from there to your home costs money too, ya know.

Airtel either owns or co-owns roughly 65%, Tata owns or co-owns about 34% and Reliance owns or co-owns roughly 11% of the international capacity. Keen mathematicians will notice that this adds up to more than 100%: Bharti & Tata in particular are in separate consortiums with other ISPs on all of their cables bar one each, which causes their % of ownership to overlap but it must be said also that in these cases, the prices are set by the consortium which includes ISPs from each country connected to the cables, which in some cases also includes governments.

Not to mention that these cables are not cheap to build - we're talking hundreds of millions of dollars each here - they don't exactly take suitcases of cash to the cable construction company and tell them to go forth - they either finance the build with large banks or if they do pay up front then they do the same as I've been talking about with everything else: the cost of building is amortized over time.

 

[naveen_reloaded]

Oh dear. Welcome to Economics 101.

I'm not talking about equipment that the users ever see, but all of these things have ongoing costs. It depends on the ISP, but even if we buy/build real-estate there are ongoing costs for it such as maintenance, taxes or mortgage payments, and even if it is a one-time cost (that is, the company pays for it up front), that cost has to be amortized over some period of time. In the case of smaller ISPs, the real-estate is rented and thus has a fixed monthly cost.

Cables may have a one time cost but as with real-estate, that cost has to be amortized, and even then, the payments to the municipality are also recurring, and even then, the cost of the actual cabling is minimal compared to the cost of the payments to the authorities and the civil works (where the burying of cables is concerned, that is), but if I spend 1000-crore on my network today, I'm expecting to get that money back over the period of say 5 or 10 years, not 1 month..

you are right ... there are maintanance cost ..etc ..

I thought VAT is supposed to do that . I mean as hte name implies Value added at each level. Doesnt VAT , we pay include the maintance / tax (doubtful) / staff charges too ?
You are right that the typical ADSL modems provided by most ISPs are dirt cheap and that at least half the cost is paid for by your deposit. This is where we in particular at a disadvantage: our modems are way not cheap, but we don't want to charge a Rs5000+ deposit!

thanks for the inside information. I knew it when i got a Netgear modem in place of BSNL`s crappy modem.



The same argument can be put forth that if the FUP is calculated correctly, most users should never reach it. As for the degraded connection, this is one of those things that you must put up with in the world of the copper (or wireless) last-mile.

Is that a vicious cycle we are setting in today ?
Lets say all ISP implement FUP. and lets say on a 512kbps speed plan (i download around 100+GB) and set FUP CAP at 50GB/month. Normal users , heavy downloaders will automatically tune to the ISP`s CAPing over a period. When more users come to know , particularly normal users , particualrly in India where literacy about internet is very poor , people will think that watching few youtube videos will exceed the CAP and will have to pay more at the end. Or if one bill lands them beyond the said price for downlaoding more than the CAPd limit... these normal users will thier internet usage far below even 25GB. ISP can again take this sample and push the FUP futrther down , stating that avg usuage is now 25GB and we are setting CAP at 30GB .. cycle continues.

Please dont tell me this wont happen. In India i know lots of people who will shift to these measures.

We heavy users are asking just one thing. You ISP know how many users are average and how many are heavy downlaoders. normal household connection( i mean to say the house hold mothers who use for few skpe chat or emails and father to see thier company emails.. etc) can have these FUP conenction and i agree , they wont cross the limit , i can even assure you that they will utilise only half of FUP CAP. My mom has a connection around 1GB/month for 250rs and i think for the last 4 months or so , she would have used 100 MB Max. (250Rs plan is the starting plan). ISP can offer such starters plan and ask these users to move accordingly based onthe usage.

On teh other spectrum , you can give heavy users , one who knows what they want from thier internet connection , full freedom and charge accordingly. As i said , we heavy users are not shying from paying a premium . I am ready to pay 1K~1.5K for 1MBPS UL ( currently reliance has this plan) even 1.5K~2K for 2MBPS UL if thats feasible


I don't disagree with you there - the way the plans are advertised in India has often felt wrong to me in the past because the FUPs were so tiny. However, with caps slowly becoming somewhat more "reasonable", this is becoming less true and again, if they are calculated correctly, then the majority of users should not reach it and in such a case only the heaviest of downloaders should be affected.

Also, the fact that you were the only one who subscribed to the Rs1,500/month plan says something too: the ISP is going to have to get a lot more subscribers paying for the Rs500 plans to break even. From there, the numbers are a juggling act as to how much is needed to supply the whole village etc.
In our case, we need something like 250 subscribers to break even on fibre in a given area, and that's at our prices: forget the sub Rs1k prices other ISPs have. Fortunately, in metros this is quite easy because of the density, and in villages it's not as bad because the costs to the municipality (etc) are less, so a larger geographical area can be covered for the same amount of money (which ends up balancing out the numbers).

same can be said in reverse. I am damn sure about 80% connection taken in my village by others use hardly 1~2GB / month. example i have stated above , so arent you guys making excess money from unused BW ?





Fair enough.
So am i right , there is no bussiness plan for startups ?



I think you didn't read what I wrote: I wrote that Rs2,200 might be some fictional cost of *delivered* bandwidth - that is, including all the equipment, staff, last mile etc. In real life the numbers are different but of course they also vary widely based on how well the ISP has negotiated with the upstream provider and how much they buy. The cost of *wholesale* bandwidth is under Rs1,000/mbit/month for the big ISPs but that only gets it as far as their offices in either Mumbai or Chennai (and it costs more for smaller ISPs). The distribution network that takes it from there to your home costs money too, ya know.


I agree . I have few things too , I have been a bsnl customer for 2+ years now , If you take dialup(landline) too , then it may well go into 5~7 years. all the way the line is same. So if you get a customer in , the intial laying cost may be high , but on a average say , a customer stays on yout network for 1-2 years , wont that cost be covered + you get profit too ? (monthly charge for these)

thats good and i think BSNL is giving a good plan 512kbps for 750+TAx ( actually it should be 500RS , but 250 for extra charges is very nominal)

Airtel either owns or co-owns roughly 65%, Tata owns or co-owns about 34% and Reliance owns or co-owns roughly 11% of the international capacity. Keen mathematicians will notice that this adds up to more than 100%: Bharti & Tata in particular are in separate consortiums with other ISPs on all of their cables bar one each, which causes their % of ownership to overlap but it must be said also that in these cases, the prices are set by the consortium which includes ISPs from each country connected to the cables, which in some cases also includes governments.

Not to mention that these cables are not cheap to build - we're talking hundreds of millions of dollars each here - they don't exactly take suitcases of cash to the cable construction company and tell them to go forth - they either finance the build with large banks or if they do pay up front then they do the same as I've been talking about with everything else: the cost of building is amortized over time

i am sorry , its not reliance ,its tata i was to mention next to airtel in owning the major network. my bad.

tell me why there is no cartel among them ? also when the server prices and hardware prices are crashing fast to dirt cheap level , why is the cost of BW increasing ? if you can pour in more info on this regard , it will be helpful.
I agree laying fibre via oceans are not cheap. but any company which have to start wont have much problem as they can raise through Public offering on market .


Many feel current practice of CAPing is very restrictive and will ultimately retard the growth of internet. ISP dont seems to care about future . As i said ,they have found a sweet spot.

I will add one more fact. Rewind 5-6 years back , internet moved from 128kbps (famous sify plans If my memory serves me right) to 256kbps , then to 512kbps , all teh while giving customers more room to breathe.

for the past 2 years or so i am stuck at 512kbps. If you call that improved of internet in india , then i have nothing to rebute your argument. If ISP`s are mainly aiming at bringing high speed internet which cant be even used properly and at the end day we customers feel dejected , what is the purpose of such improvement ?

 

[mgcarley]

you are right ... there are maintanance cost ..etc ..

I thought VAT is supposed to do that . I mean as hte name implies Value added at each level. Doesnt VAT , we pay include the maintance / tax (doubtful) / staff charges too ?

VAT (Value Added Tax) is a government tax on anything a company sells. The value is value as in "invoice amount" not value in any other context. ISPs can't benefit from that because that VAT gets passed on to the government (less any VAT paid on things the ISP buys).

thanks for the inside information. I knew it when i got a Netgear modem in place of BSNL`s crappy modem.

It's hardly inside information - it happens all around the world. ISPs give equipment that will get the user connected - they don't really need to care about the quality of said equipment (wrt throughput and "extra features") so long as it outlives it's warranty period.

Is that a vicious cycle we are setting in today ?

A little bit.

Lets say all ISP implement FUP. and lets say on a 512kbps speed plan (i download around 100+GB) and set FUP CAP at 50GB/month. Normal users , heavy downloaders will automatically tune to the ISP`s CAPing over a period. When more users come to know , particularly normal users , particualrly in India where literacy about internet is very poor , people will think that watching few youtube videos will exceed the CAP and will have to pay more at the end. Or if one bill lands them beyond the said price for downlaoding more than the CAPd limit... these normal users will thier internet usage far below even 25GB. ISP can again take this sample and push the FUP futrther down , stating that avg usuage is now 25GB and we are setting CAP at 30GB .. cycle continues.

They could only really do that if that's what the actual statistics are, and unfortunately statistics are notoriously inaccurate when you take an average across 1+ lakh users (even 10 lakhs for the larger ISPS) - there are an awful lot of users on all plans that only download say 2, 5, 10 or 20GB per month, and that's fine: they know how much they are going to pay and they're comfortable that there will be no extra expenditure on Internet. What most ISPs do NOT have at the moment is plans with FUPs of 200, 500 or even 1000GB. Why not? Such plans exist in Australia, so it must be worth doing.

That's why our FUP is based more on the price than the speed of the service and given the speeds we're offering we can safely expect users to download what other ISPs might consider "hefty amounts", but we are fortunate that our higher starting prices and lack of

 

[naveen_reloaded]

I really appreciate you for taking time and replying to each point.
quoting each para again will by itself will become a thread

The last few para`s is all i was looking for and thanks for agreeing to it (sweetspot).

Unless we customers dont ask for more bandwidth and better speed , ISP`s in India and all around the world can easily make money what i would call a "Yesteryear`s Technology" . Just like Firefox came into action when IE was dictating terms on how a website should be created to look good in IE.

Only problem for customers like us in rural area are , your ISP and ISP like Beam will take atleast 2-4 years to each us. Not to mention even BSNL which is the only ISP here dont provide 2 mbps . maximum is 1 mbps. So stopping development and cutting on infrastructure and not actually upgrading to better equipments will only delay penetration in our areas.

again thanks a lot for that detailed reply.

I wish if you can comment or tell us more on the national gateway level CARTEL. Do you think taking it to media level will result in anything ?

----------

one more request . Can you tell us what equipments needed for running ISP ? Just the main ones ..

 

[mgcarley]

Only problem for customers like us in rural area are , your ISP and ISP like Beam will take atleast 2-4 years to each us. Not to mention even BSNL which is the only ISP here dont provide 2 mbps . maximum is 1 mbps. So stopping development and cutting on infrastructure and not actually upgrading to better equipments will only delay penetration in our areas.

It would help to know where you are. I've stated in my own threads that areas where there is only the single player (usually BSNL) are attractive: firstly because they have no other competition and secondly often they would be quite easy to deploy in. It could take as few as ~250 subscribers to break even.

I wish if you can comment or tell us more on the national gateway level CARTEL.


What can I say, really. They don't seem to co-operate with each other. They do compete with each other - when I first started talking to the providers I got the price of an STM-1 down from 2Cr to just over 1Cr a year in the space of about 3 weeks (this was in early 2009).

Do you think taking it to media level will result in anything ?

Not really. From what I can personally ascertain, it's due more to government policy than anything else. There are companies dying to build cables to India but they just get stymied at every turn.

one more request . Can you tell us what equipments needed for running ISP ? Just the main ones ..

It varies widely and depends on the size... that's kind of like asking "how long is a piece of string". It could be anything from a couple of servers & some wifi gear to 1 or more rooms full of servers, routers, switches, patch panels, ups' and other bits and pieces.

Here's medium-sized Australian ISP Exetel's POP & Core diagrams from a few years back:





----------

Here's another ISPs info (not very detailed, but maybe helpful nevertheless):

http://www.iinet.net.au/iinetwork/dslam-installation.html
http://www.iinet.net.au/iinetwork/network-diagrams.html

 

[naveen_reloaded]

for my location to be exact here it is [noparse]http://tinyurl.com/mecheri[/noparse] [/B]There are potential buyers , few colleges in and around , many schools and good number of shops , As like all other companies you would do research on a place before stepping in right ?. I will just add one thing .. me and my friends tried asking airtel which is providing service in nearby city to expand here to our place say 6 years before. Still nothing!! As for the goverment policy thingy , what exactly is it ? I mean looking at the 3G scams , i wonder why this CONGRASS wouldnt let companies like you to lay fibres (international)?As far equipments , i just want to know the speed at which that segment is improving in terms of performance and price.. the pictures are giving me headspin .

 

[mgcarley]

for my location to be exact here it is [noparse]http://tinyurl.com/mecheri[/noparse]

Always in Tamil Nadu... I was just having a discussion just the other day with someone in Gingee (which has a similar population to your town) about providing services in small towns.

There are potential buyers , few colleges in and around , many schools and good number of shops , As like all other companies you would do research on a place before stepping in right ?. I will just add one thing .. me and my friends tried asking airtel which is providing service in nearby city to expand here to our place say 6 years before. Still nothing!!

It is regretful that towns of this size are basically ignored by the private players, but probably it's a simple question of justification.

In a small town - especially one without a railway station as both yours and Gingee are lacking - it might be such that the private player would need to rent some fiber from BSNL and at the prices I've seen them offering domestic capacity, that can really kill any hope of a private player that doesn't have it's own fiber nearby entering that market.

As for the goverment policy thingy , what exactly is it ? I mean looking at the 3G scams , i wonder why this CONGRASS wouldnt let companies like you to lay fibres (international)?

The usual claim is security. There has also been fights among the big boys with regards to cable stations - VSNL, being formerly a government entity, has had the sole rights to own/operate the cable stations for years but Reliance at the very least has contested this quite loudly when they purchased FLAG... As I recall, it's still a big mess.

As far equipments , i just want to know the speed at which that segment is improving in terms of performance and price.. the pictures are giving me headspin .

This is one area where brand is quite important, as there can be a huge variation depending on what equipment is used and what brand it is.

At a large ISP, you could be looking at $500,000+ each for the mac-daddy routers from Cisco, Juniper, Alcatel or Zhone (can be less for other brands) in the core, then at a PoP you could be looking at anything between $1,000 and $200,000 for the terminating equipment (depending if it's wireless or DSL or ethernet or cable) and again, all this can vary widely by brand, but like I said, the question is still quite open-ended and difficult to give any kind of definite answer to.

Higher cost doesn't always translate to higher quality/better performance - the difference between a $400k router and a $500k router might not be significant - but at the same time if you buy the cheapest equipment you are almost guaranteed to get what you pay for. This is most noticeable (especially to the end users) when companies have used brands like Teracom for their WiFi services - it may work, but the performance is poor, pings are high and the service becomes slow when even a dozen users are using the network. Even I made the mistake of trying a cheaper brand for our Hayai Lite service in NZ before settling on what we have now, but I've learned my lesson.

In some cases a poorly performing ISP could have several reasons for being so:
1. the ISP is running equipment that is of poor quality (in the core this isn't usually the case - most ISPs choose to run Cisco at the core which is usually decent performing stuff and is almost the defacto choice, but towards the edge of the network - where the users are - sometimes the equipment quality begins to decline).
2. the ISP is running their links at too high capacity. In India the TRAI recommends a maximum of 90% CBR (constant bit rate) before the link is considered saturated and needs upgrading (we work on the assumption of 75% = saturated because it would be very easy to spike from 75% to 90%).
3. the ISP simply has not purchased routers that are capable enough of handling the traffic volumes they do and as a result the performance of their services can suffer.

 

[Zaman2011]

Like others have said, the analysis of the situation is completely off. It should be very clear that residential broadband is oversold (worldwide), and with good reason. You've heard of the "contention ratio" - that's what that means. For example, in India every 1mbit/s we buy, we can sell a maximum of 50x 1mbit/s connections (oversimplifying it, but that's the jist).

Now for the reasoning: cost. Sure, ISPs buy bucketloads of bandwidth, but riddle me this: ISP-X currently has lit up capacity of about maybe 200gbit/s (or less) - for 2 million customers, that's essentially 0.1mbit/s per customer. Even if they're paying say Rs1,000 per mbit/s (it's less than that, but let's keep the numbers clean), that doesn't factor in the cost of staff, real-estate, taxes or the equipment needed to handle all that bandwidth, let alone the last mile. Let's be generous and say the cost is then about Rs2,200 per mbit/s delivered to the home after everything. Add a little profit margin (25%) and you've got just short of Rs3,000.

Now, how many people are willing to pay Rs3,000+ a month for 1mbit/s? Not many. So how is ISP-X going to be an attractive option to it's customers? Offer say 4mbit/s for about Rs1500. But oh no! If that 4mbit/s is uncontended, they're going to lose Rs6,300 per customer per month at that price.

Enter the Fair Usage Policy.

4mbit/s allows roughly 1,200GB a month of data transfer. For ISP-X to break even, they can have a maximum of 300GB per user if the contention ratio is 1:4. But ISP-X doesn't want to break even, they want to make a profit. With contention ratios of between 1:12 and 1:20 is common in India, the indirect result basically causes an FUP of between 60 & 100GB per month. For 1:12, this means revenue of say Rs18,000 minus costs of Rs8,000 minus any other costs that may be incurred such as subsidizing modems, maintenance, DoT licence fees and in many cases, bribes or revenue sharing - the margins are still only between 30 and 50%.

The situation is slightly different in the US, and given that you've got a line that's "up to" 24mbit/s, the numbers actually get better (not by much, but the premise is kind of what we're working on) - where although bandwidth and peering is cheaper, I don't think the last-mile is, and staffing definitely isn't: then you've still got a cost of say $15 per megabit delivered (*24mbit/s = $360), contention ratio of say 15 = $750 of revenue less $360 of direct costs less incidentals leaving similar margins and similar FUPs which are dollar to gigabyte a bit similar to India. Again, oversimplifying it here (and perhaps in the ISPs favour, especially in this example), but I hope the point is made: bandwidth by itself only makes up a relatively small fraction of the total cost of supplying these services.



Residential broadband is NOT designed for anyone to start their own hosting company, and in many cases it's even against the T&C. Business connections are a bit better in this respect if you're willing to pay for them, but there are plenty of

 

[mgcarley]

For example, I have a low end dedicated server in Europe just for fun which costs me less than 40 USD and provides me unlimited bandwidth on a 1Gbps shared line and with FUP of 5000 GB after which it reduces to a 10 Mbps line. So, is the residential bandwidth in Europe so less for the providers to allow such massive bandwidth usage at low costs. I mean for a mere

 

[Zaman2011]

Yes. There are significant price differences between Europe & India, the US & India and to a lesser extent Europe & the US, and there are several different things at work here.

Firstly, peering in Europe is dirt cheap.
Secondly, in almost all cases the last-mile lines are laid by the government or incumbent service provider for use by any & all service providers and because *everyone* is using that same infrastructure, whoever owns the lines gets paid no matter which provider the user is actually with.
Thirdly, peering in Europe is dirt cheap.
Fourth, in many cases a significant amount of traffic is staying within Europe because either content being accessed in Europe is hosted in Europe itself OR the major sites all have CDN nodes in the peering exchanges which causes said traffic to remain in Europe even if it comes from the US or Asia or elsewhere.
Fifth, peering in Europe is dirt cheap.
Sixth, many Europeans are used to not being deprived of bandwidth, and as such many users usage remains at manageable levels.
Finally, even transcontinental lines are dirt cheap (especially US-Europe routes).

I've was in Romania in 2008 and even at that time I found the Internet access (and mobile access for that matter) to be surprisingly good - even outside of the capital. Their government has been doing a lot of good work in that respect - so much so that Constanta now tops the speedtest records for the European continent.

They are a little bit lucky though because of their geographic location - a stones throw from Germany (via Hungary) whose telecoms are doing great things, and France, who is not only home to Alcatel Lucent, but who currently have some of the busiest Internet routes in the world, with average bitrates of over 2Tbit/s. Furthermore, they're "on the way" to these places from Turkey, Bulgaria, Greece, Moldova, Georgia, Azerbaijan, Armenia & Iran (though some ISPs from that region also partially route over Telecom Italia for the sake of diversity).

Likewise the access is quite good in the Ukraine (mostly in cities though), although some Ukrainian ISPs have a slightly interesting pricing scheme: you pay some amount per month as you would expect, but your access speed is different for domestic traffic (usually 10 or 100mbit/s) and international traffic (as low as 1mbit/s unlimited for US$6 per month, but 100mbit/s is something like US$50 or 60). It used to be similar in NZ but that was a few years ago.


Thanks for explaining. Makes a lot of sense especially the domestic and regional usage part. I have noticed this thing in the whole of Europe that most European websites are made in their regional language instead of English and so automatically attract local people. For example, websites in France normally use French for their websites, German websites use German language and target local market, Russian websites are mostly in Russian including top sites such as Rambler, Spain and Portugal also targets local audience (though they target Latin America to an extent too due to the same local language across the continents) and so on. It is quite different from countries such as India where all websites are in English and mostly hosted in the US. Even for normal sites, Indians would probably search for a English website located in US or US rather than for a Hindi or a local dialect website. Most Indian webmasters also do not target the Hindi audience or local audience as Indian online market is not the best and they also do not host here as hosting in India is also a lot more expensive than US etc. So, it makes sense why Europe does not need to spend up too much trans-continental bandwidth and in many cases trans-country bandwidth as most European websites are hosted in their local servers itself.


In short, the percentage of users who use 60-70TB a month on these servers is very, very small. Most users would struggle to use even 100GB, let alone 1, 10 or 100TB. Basically about 1,000 customers at $30 each is enough to break even on a 10gbit/s international line from NYC to Europe (though as per previous examples probably 2,000 customers to pay for all the other incidentals like staff and whatnot).

Furthermore, most of this bandwidth is being served to domestic audiences - and peering which is dirt cheap even compared to a line of this size. Going by the price of a 10gbit/s peering line in Europe where you're looking at 1,200EUR per month tops, and it's even cheaper in the US. So then we see that we've got, say, US$64,000 worth of bandwidth costs per month (assuming we only had two peering lines & two international lines) which can be taken care of with just ~4k customers - compared to India which is quite significantly more.

As we know, companies like 100TB & Softlayer (who have multiple data centres as far as I know) and the various hosting providers in Europe have WAY more than 4k customers, so it becomes a simple matter of volume versus usage: they have huge volumes of customers who hardly use anything close to the amount of bandwidth they're actually purchasing, so while the lines are vastly oversold, they're oversold in such a way that it doesn't really affect anyone.

This situation is, however, very different to residential broadband in that even a $50m data center is peanuts compared to what an ISP can potentially spend on infrastructure. Moreover, a larger percentage of residential broadband customers are far more likely to actually *use* the bandwidth that is allocated to them compared to a hosting customer.


I agree but for somebody like 100tb it is a bit different as they are MAINLY used by BANDWIDTH HOGGERS only. So, a person who spends 100GB would probably never use them as they can go for a shared hosting, VPS or a low end dedicated server at an EXCELLENT dedicated server provider like Softlayer instead of a not-so-great customer support high bandwidth provider such as 100tb. They can get it for a lot cheaper and also get a lot better quality with some better provider if they do not need such massive bandwidth and have normal websites. But for the most part it makes sense and thanks for explaining.