You are correct in that the speeds are "up to X mbit/s" in the marketing materials and the T&C contain no guarantees of speed/uptime/etc, and that X mbit/s is in no way achievable to the whole wide Internet, but this is a service served from local-ish servers in a country where peering and transit is - especially relative to certain parts of the world - quite cheap and fairly readily available.
ISPs have a responsibility not only to their stakeholders but to their customers to provide something resembling the service being offered. If I pay for a 50mb connection and I can't even stream something at 5mbit/s reliably or at all, that seems to be a problem that only the ISP can solve, and I would go so far to suggest that if the ISP could be breaching their contract if the service is unreasonably less than what they are selling.
I would even think that if said stream works properly when the subscriber is using a VPN but not when using the ISPs standard routes & DNS, that ISP is more or less responsible for rectifying that situation, that is to say, making the VPN less necessary, because then the "best effort" (as most contracts define residential service to be) isn't really "best effort" at all.
It's a given that ISP should have different pipes to different upstream providers - in the US I can reasonably expect a large ISP to have multiple 10gb links to Cogent (due to Netflix) and Level3 (and maybe others), but maybe only a couple gigE links to Telecom Italia or Tata - it all depends where the customers are drawing the content from, and given that in the US Netflix is a rather significant content provider, it should make sense for all sorts of reasons that the ISP should do what it can to ensure it's connectivity to Netflix is halfway decent - buy more transit, peer directly or cache, whichever makes the most financial sense. But do nothing? Not really an option. Especially when they start introducing new tiers or upgrading customer speeds (as my ISP did just this week).
I don't expect to get my full speed to every server in my state, let alone the country or the rest of the world, but there's a lower limit to what I expect as reasonable too, which goes back to that whole thing of getting something even vaguely resembling what I'm paying for, and getting 5% of the plan speed from a domestic source is not really acceptable to me (as a consumer) - if I, in the midwest, can download from a server in India at 5-25mbit/s (based on my own testing, speeds can vary a lot by city), why can't I get Netflix in HD from a facility NJ, GA, CA or TX, which only takes about 5 or 6mbit/s?
By the way, Shakrai, I don't know which part of the country you're in, but $1,100 for 30mb stinks of a ripoff. For that price (including tax) in a not-very-large town (30k-ish), I get more than 3x as much bandwidth (dedicated active fiber with SLA, including the last mile, private/non municipal, gig line which is rate limited at the CPE so speed upgrades take about 90 seconds... but my burst-rates are more than your total BW). It's from a Tier 2 CLEC but the upstream is L3 and Qwest and both Netflix and Youtube work a treat in HD & 4K.
"Completely trivial" != zero, and it's not Verizon's responsibility to subsidize Netflix's business model.
Apparently it's not enough for them to get free bandwidth, when the entire history of settlement free peering has relied on balanced traffic ratios. No, on top of that they want someone else to pay for their rack space, electricity, physical security, and all the other overhead that comes with running a modern data center. Essentially Netflix just wants to take your money and send it directly to Hollywood (after taking a cut for themselves of course) without any of that pesky overhead otherwise known as "running a business".
Well hang on a minute - if I have to put a rack in and provide a standard amount of power to it, that's going to cost me... let's say a few thousand a year at each location (for a large provider, let's go with around 8 locations). So let's take $17,000/year per rack * 8 DCs = $136,000/year as the cost for a major player to run 8 racks of caching servers in existing data-centres around the country.
Even with transit bandwidth being as cheap as it is (in major centres at least), for a 10gig link I'm going to be paying (as an ISP)... probably a few thousand a month, but I'm at their mercy when either end of that link (whether their fault or not) inevitably becomes congested. So taking HE.net's advertised pricing of $0.46/mbit * 10gigs = $4,600/month * 8 locations = $36,800 * 12 months = $441,600.
But wait! We're talking about settlement-free peering! Well, with the peering I'm paying for optics and rack-space anyway - though not as much as I'm paying to run a caching box, I'll grant you. The question then is: private or public? Using Coresite public peering as an example we're looking at a mere $6,000 a year for a 10gig link * 8 locations, so only $48,000 or so, or about 1/3 as much as a rack of caching servers. For private peering... who knows what Verizon's (or Comcasts or anyone elses) costs are.
While the figures are merely an educated guess based on prices I've been able to find in Chicago in public facilities which may bear absolutely no resemblance to Verizon/Comcast/etc costs for the same, so far, running a rack full of caching servers is better than transit, but peering is better than running a rack full of servers.
But that then begs another question - the customer getting what they're paying for. If it weren't for all the anti-class-action clauses in ISP contracts, they'd probably be getting sued a lot, and the TV would be full of commercials for lawyers suing ISPs instead of (or more likely, as well as, pharmaceutical companies)... but generally speaking, most ISP customers in the US are getting ripped off big-time. I myself just got a 50mb cable connection (I was on 100mb fiber, but I moved), and I can get 50 (even up to 70mb) on speedtest.net but Youtube won't give me 1080p or 4k on anything (it doesn't stutter or buffer, it simply doesn't offer me the option, even on videos I know to have it - the best I can get is 720p), and Netflix seems to only run at SD. So I'm marginally annoyed about that.
And if it's such a small cost to upgrade the speeds of the peering links, why aren't they being upgraded? And if peering can no longer be settlement-free due to the asymmetrical nature of the traffic, why can't the ISP combine the two most cost-effective options both to improve the customer experience AND hopefully alleviate some of the asymmetry.
Is $136,000-ish a year really going to hurt that much? That's only... what... a couple of thousand extra customers that will need to be signed up to break even on that investment (assuming about $70/mo/sub revenue less expenses). Plus, with caching servers, I could even use that to attract more customers because sites like this one will have people saying things like "choose this ISP if you want to watch Netflix because caching" (probably... that may also not happen, who knows)...
In the grand scheme of things, it's barely a drop in the ocean of expe
Whenever I've gone to a store looking for a phone, I make it a point to avoid the iSection because I don't want to get any on me.
Last time (less than 3 weeks ago, as it happens) a $CARRIER employee asked "Why not get an iPhone?" when I was clearly looking at Androids, I told him I couldn't take him seriously and waited for another employee who would sell me what I wanted.
So no, I don't believe it was a mistake for a large percentage of people - the customers were probably (in their minds) "graduating" to the top-of-the-line flagship carrier device after having had a feature phone, a low-end smartphone (probably Android) and then a higher end smartphone (probably a better Android) every 2-ish years for the last few contract terms... the kind of people with not a great deal of money to spend so they kind of have to prove to themselves that yes, they can handle having a $700 smartphone (but who still buy the el-cheapo sub-$500 laptops at Wal-Mart and load them up with more malware than I can shake a USB-stick at).
Were they **really** satisfied? Well, chances are the marketing survey was taken while the phone was new, so, in comparison to their old phone, yes, it's going to score 4 or 5 stars. Ask the same people half-way in to their contracts and the results are probably not nearly as pleasing.
Depending on the ISP(s), this type of behaviour might be prohibited by the ISP(s) that are providing services at each of the complexes, as would charging individual tenants for the access and numerous other activities.
Typically even in a dedicated line most providers are OK to let you provide access on a straightforward "as-is as part of the rent" basis but they might interpret such meddling and the additional revenue generated as charging for a service that isn't technically theirs to charge for; so the company in question might want to check the T&C in each of their upstream contracts (eg does the contract permit reselling of bandwidth) before their upstream provider(s) lawyer up and sue for breach of contract.
If, on the other hand, the company has assets like it's own ASN/IP space/FCC registration (and purchases from a non-competing provider such as like Level3 with permission to resell the bandwidth in the contract), then this whole thing might be a different story.
(This information is based on my own experiences running dedicated fiber lines from other some large ISPs and provide access to similar complexes in college towns in a few different states and discussions with various providers in each state).
Lets put this all into perspective. Transit is priced about $0.45 retail, and costs even less. But lets give Comcast a bit of wiggle room and assume that on average, bandwidth costs $0.50/mbit. This is bandwidth that can reach anywhere in the world. Comcast oversubscribes about 20:1, which means that bandwidth is really about $0.025/mbit per customer. Comcast then turns around and charges $100/month for 100mbit, which only costs them about $2.5 in bandwidth. Then Comcast complains that Netflix is using too much of this bandwidth, because Comcast's network can't handle providing a measly 5mbit/s to many of their customers at the same time.
If you buy from the right companies in the right quantities in the right markets, yes, that price is accurate. Outside of those markets (smaller towns) with smaller quantities (n*1gb instead of n*10gb or n*40gb) leads to a higher price as averaged across all service areas (plus the cost of the middle-mile). Although in Comcast's case, I doubt the overall rate per megabit would be significantly higher.
Moreover, that bandwidth does not reach anywhere in the world - transit can cost per route kilometer or per megabit (or some combination of both), so they might buy a connection to London which costs say $3/mb but a connection to Hong Kong which costs say $4/mb, and if the destination is in (for example) Singapore, then the traffic is likely to be sent via London unless packets are being dropped, even though sending it via London might be less efficient from a distance perspective. Or in the case of something a bit more domestic, this might mean configuring things such that all Netflix-originated traffic may only come in via the cheaper Cogent links even though the more expensive L3 links might have spare capacity on them - to the detriment of the Netflix service and Netflix customers experience because as far as Comcast is concerned, who gives a crap about the customer.
As far as the costs are concerned, using the same math ($0.50/mb) and contention ratio (20:1) you come up with a cost/mb of $0.025/customer/mb (personally I suspect their contention ratios are WAY higher than that based on the fact that their links are allegedly permanently oversubscribed, but let's roll with it) but strictly speaking, this isn't exactly true - it's not compression - you're not going to magically be able to transfer 6TB/month of data over a link that can in fact only do 300GB/month (I use 300GB as a nice round figure for data transfer per megabit per month - it's a little lower than the theoretical maximum but accounts for network overheads).
Really, the COST doesn't change, just the actual amount of available megabits to the outside world available to each customer across the customer base does, and this is one of the reasons for data caps being implemented (so it would be better to count a per GB cost than a per mbit cost since we're looking at the volume of traffic, not the cost per subscriber mbit as this is only ever going to be measured by Comcast up to their nearest PoP/node/DC). Using a 300GB cap or 1mbit as an approximation, it could be suggested that out of their $100 Internet bill, the average subscriber is paying a grand total of $0.50 for bandwidth (other parts of the infrastructure, CAPEX/OPEX, overheads etc notwithstanding, of course). Which is pretty pathetic, really, and my humble opinion is that now that Netflix and Comcast are directly connected, traffic between those networks should no longer contribute to any data cap that may be in place.
The problem is in part that even though their last mile should theoretically handle a bunch of 5, 10 or even 50mb users all using the network at the same time, and if measured to the nearest Comcast server there's no reason any given subscriber shouldn't get most or all of their subscribed speed most or all of the time... but since we're trying to access the WORLD wide web, that's not how the customer sees it, and as the middle and/or first miles are less able to handle all those subscriber
Is it worth noting that it's something like $2,000 per premises passed? Even if it were only $700, this difference does change the economics substantially... if you're the 5th person on a block and the first 4 don't also subscribe to the service, they've effectively shelled out $10,000 (or $3,500) to supply one customer rather than the figures often quoted. Of course, if all 5 premises subscribe, the economics are favourable.
Which is why open utility/municipal/government last miles make sense: since the bulk of the cost comes from the roll-out, you've got 5 premises who will be subscribing - even if they are subscribing to 5 different ISPs - and the return on investment is being made on 5 * $2,000 rather than 1 * $10,000 (or 5 * $700 rather than 1 * $3,500 if you prefer) meaning everybody's bills will be lower because the costs can be spread out across more subscribers.
We've heard this argument 1,000 times before but, if you'll excuse the expression, it doesn't hold any water.
Having been to most of Europe and most of Asia, population density in cities on those continents is sufficiently similar to cities in the US so there is no excuse for Internet access in any American city to be any worse than it is in any other city on any other continent, and there's plenty of dark fiber around, so with routing equipment being as cheap as it is now (even at the 10gbit/s level) it's not a question of backbone connectivity either.
Rural connectivity - that is, Internet access to farms or sparsely populated areas - is certainly an issue to be tackled, but there's no reason even small-ish cities can't be provided with decent access.
But we can't compare apples to oranges, either. You can't compare a 100/100 pipe in central Stockholm to a 0.75/0.25 pipe in the middle of Montana. After all, you might be just as likely to get gigabit Internet in rural SK/Japan/Sweden as you are in rural USA unless it's regulated that the ISP or infrastructure provider has to do it even at a loss.
What *IS* different about all those countries though is that the last mile is regulated to be open and shared among competing providers, so they keep their customers by providing better services and faster speeds rather than by protecting their monopolies or suing the competition out of existence and this makes for a fairly significant difference in availability of services.
Amazon can make a profit on 2c/GB because of 1. How much traffic it buys overall and 2. The locations of it's data centres - they're usually in or near fairly major places with IXs to plug in to and cheap transit (and if it isn't cheap at first, it becomes cheap because of #1).
The same is not necessarily true for ISPs. As much as I too dislike Comcast and other such providers and what Comcast in particular is doing, the cost of delivering bandwidth where I am at the moment (Southern Illinois) is not necessarily the same as the cost for delivering bandwidth in, say, Chicago, where I could probably get that cheap $0.46/mbit transit from Hurricane Electric if I needed it there and/or had dark fiber from there to here, but, based on my costs, my per-GB rate down here breaks down to about $0.035 before overheads/taxes so by basic rules of accounting I need to charge at least $0.10/GB to be profitable (which, mind you, is still more reasonable than the overage rates from the likes of Comcast and Mediacom which are $10/50GB or $0.20/GB).
Plus an infrastructure fee - which Amazon doesn't need to charge (again, in part, due to volume).
To put it in perspective, Frontier only has a 2gbit/s pipe to their nearest CO (as of earlier this year), so I'd imagine their costs aren't a whole lot cheaper than mine here (of course, I could be wrong) but in some of their other markets it may be far cheaper, so, IF my services were nationwide (or at least, significantly more widespread as Comcast/Frontier/Mediacom) I'd be taking an average cost of delivery and basing my retail prices on that.
Of course, being that I'm hopefully not an idiot, I'd definitely take a Netflix box in my network if they wanted to give me one. I'm not too concerned about it just yet though - our connections are only 50-60% used during peak hours at the moment.
Only 2.6 seconds? I've used Internet services on more than one continent that have had things so badly configured my ping times - even to servers like 8.8.8.8 - were more than a minute. Even recently in the US I was using something that gave me pings of 13 or 14 seconds to Google or OpenDNS or something.
For example, on our most widely used product in India there's a fairly nominal charge of Rs199 plus a flat charge of Rs10 per GB, in the US for a similar product it would (usually) be $20 plus $0.15 per GB. 30GB minimum purchases apply but subscribers can purchase more usage half-way through the month if they want to.
And there's no speed limit on usage-based plans (usually the subscribers port speed is 100mbits, but in some areas we're going to a gig) - I figure when it comes to speed and usage, we should limit either one or the other, not both. At least not artificially.
What if you're like millions of people both in this country (I assume you refer to the US) and others who are tax-payers but not citizens? Like myself. I pay tax in more than one country that I am not a citizen of... so unless you're suggesting non-citizens can get out of paying tax...
Ergo, really, taxpayer really does seem like a more appropriate term when it boils down to it, because citizenship comes with certain rights that aren't afforded to non-citizens (so, you're right about the "weight" thing) but when it comes to the funding of government through taxes, it makes no difference whether the individual is a citizen or not because if all else is equal, the tax burdens of either are equal and their taxes are spent the same way on the same stuff.
There are lots of such routers now. Ubiquiti EdgeMax Lite ($99) springs to mind but there are a bunch of others as well.
Granted, configuration may not be as straightforward as your average home router and you'll end up with 4 devices for 2 connections (1 router, 1 wifi, 2 modems) but it has an half-way decent GUI and a wizard so it shouldn't be out of the question for someone who is at least partially competent to set it up for dual-wan.
Also, I think the Zyxel modems that Frontier give out have dual-ADSL ports in them but they're not very good IMO. I don't think they're terribly expensive though.
I don't know where you are but "several thousand" sounds like AT&T/Verizon money to me - I've been quoted over $5k for 100mb both here (Southern IL) and in other states while the local CLEC charges a mere $1,000/month for 100mbit symmetrical / dedicated fiber (not 95th percentile) with all the usual things like a/29 and 4 hour MTTR - (it's about $750 for 50mbits and $400 for 10mbit/s I think)... once the packets hit St Louis we're on Level 3 and Qwest which means most stuff people are accessing is pretty speedy and there doesn't seem to be any congestion.
And unlike AT&T and Verizon, the more you buy, the cheaper $/mb and they aren't using crappy bottom-end Cisco routers.
$10/mbit works out at a cost of $0.035-ish per GB of data and while I'm not talking middle of nowhere, I'm not talking about a large market either.
I think your numbers are a bit off, both as far as the price of hardware vs ROI and as far as the "huge profits" are concerned - aren't Comcast and Verizon supposedly making 40-90% profit on multiple Billion dollar revenues? (Granted, for these companies this may be an overall company margin including all business units, not necessarily the margin on HSI).
In any case, my own companies are much smaller than them (so higher cost per customer for equipment and more expensive bandwidth because I'm not buying nearly as much as they are so I don't get quite the same volume discounts) and if it weren't for our expansion projects, we'd be doing pretty tidily.
And assuming people aren't dicks about bandwidth usage (eg hogging all of their bandwidth all of the time), we can even make residential services a little bit profitable without cross-subsidizing between plans (eg higher tiers aren't subsidizing lower tiers and vice versa). In the US, we'd be looking at something around 80% profit before equipment & overheads, and with a PON setup working out to under $1,000 per port (assuming brand-name gear, 10gb optics in the back and GPON for the customer-facing ports), our RoI really isn't as long as one might expect.
Population density per square mile as averaged over the entire country is really not relevant. Someone living in the sticks is going to have limited connectivity options whether they're in the UK, USA or most other countries (even our favourite poster-children like S Korea and Japan).
A town/city with 2k or 20k or 200k or 2m people is a city with 2k or 20k or 200k or 2m people in either country and, geological differences and labour costs notwithstanding,//should// cost approximately the same amount of money to service (and thus, retail prices should be pretty similar).
I put to you that if 2 last-mile projects were simultaneously undertaken to provide 1,000 customers in 10 a square mile area, but one was allowed a monopoly on it's last mile and one was not and had to be open/unbundled and had 5 operators competing for those customers as detailed by cid=46867081, the latter would have far cheaper end-user pricing, irrespective of which country/ies we were talking about.
In many European countries, food & housing is subsidized. 0.80 euros a meal (buffet style) at the University cafeteria and about 350 euros a month to live in student housing (1 bedroom apartment).
And this was in the mid-2000s, in Finland - one of the more expensive countries to live in - which meant that the Internet connection (which was included as part of the rent along with other utilities) was better even then, than I could get now in 2014 in the US (if I were a regular Internet subscriber, that is).
And car? WTF would you need a car for in most European cities? Students in most European cities would not consider a car to be a necessary item (nor would many adults) and since public transport costed about 40 or 50 euros a month for a pass that got me on any form of Helsinki public transport (bus, local train, tram) which ran several times an hour, a car simply wouldn't have been worth it.
And having also lived elsewhere in Europe, I seem to recall that situation being pretty similar in many other parts of the continent too.
In many countries (especially in Europe) modems/cable boxes/etc are included in the price of the service - I don't seem to recall being charged extra for ISP-supplied equipment in France, Japan, Finland, Georgia, Ukraine or New Zealand**, but either perhaps I didn't notice or it's simply built in to the price over the term of the contract (that is, it's an implied overhead so it's accounted for by locking you in for 12-24 months)... but typically you're on a term contract in the US as well so there's no difference there anyway.
And I seem to remember Free & Neuf boxes being valued at something in the vicinity of 200 euros at one time as well, so they're no cheaper.
**May also vary by ISP but may also be entirely wrong about the whole situation in some or all countries - correct me if I'm wrong.
Especially when "tips", as it turns out, are supposed to legally make up the actual hourly rate paid to the server to meet minimum wage, for example, the server's hourly pay is $2.50, it is expected they will make at least $6 in tips per hour causing their total hourly rate to equal the $8.50 minimum wage (numbers loosely based on Illinois, I don't know specific wage rates for other states)... so if customers stiff the server on their tips, the server gets screwed.
As a non-American though, I agree with you: tips (and prices being advertised without taxes, for that matter) are bloody annoying - I used to get caught out all the time when I first came to the US. I'd hate to be in the service industry here.
Slashdot Mirror
You are correct in that the speeds are "up to X mbit/s" in the marketing materials and the T&C contain no guarantees of speed/uptime/etc, and that X mbit/s is in no way achievable to the whole wide Internet, but this is a service served from local-ish servers in a country where peering and transit is - especially relative to certain parts of the world - quite cheap and fairly readily available.
ISPs have a responsibility not only to their stakeholders but to their customers to provide something resembling the service being offered. If I pay for a 50mb connection and I can't even stream something at 5mbit/s reliably or at all, that seems to be a problem that only the ISP can solve, and I would go so far to suggest that if the ISP could be breaching their contract if the service is unreasonably less than what they are selling.
I would even think that if said stream works properly when the subscriber is using a VPN but not when using the ISPs standard routes & DNS, that ISP is more or less responsible for rectifying that situation, that is to say, making the VPN less necessary, because then the "best effort" (as most contracts define residential service to be) isn't really "best effort" at all.
It's a given that ISP should have different pipes to different upstream providers - in the US I can reasonably expect a large ISP to have multiple 10gb links to Cogent (due to Netflix) and Level3 (and maybe others), but maybe only a couple gigE links to Telecom Italia or Tata - it all depends where the customers are drawing the content from, and given that in the US Netflix is a rather significant content provider, it should make sense for all sorts of reasons that the ISP should do what it can to ensure it's connectivity to Netflix is halfway decent - buy more transit, peer directly or cache, whichever makes the most financial sense. But do nothing? Not really an option. Especially when they start introducing new tiers or upgrading customer speeds (as my ISP did just this week).
I don't expect to get my full speed to every server in my state, let alone the country or the rest of the world, but there's a lower limit to what I expect as reasonable too, which goes back to that whole thing of getting something even vaguely resembling what I'm paying for, and getting 5% of the plan speed from a domestic source is not really acceptable to me (as a consumer) - if I, in the midwest, can download from a server in India at 5-25mbit/s (based on my own testing, speeds can vary a lot by city), why can't I get Netflix in HD from a facility NJ, GA, CA or TX, which only takes about 5 or 6mbit/s?
By the way, Shakrai, I don't know which part of the country you're in, but $1,100 for 30mb stinks of a ripoff. For that price (including tax) in a not-very-large town (30k-ish), I get more than 3x as much bandwidth (dedicated active fiber with SLA, including the last mile, private/non municipal, gig line which is rate limited at the CPE so speed upgrades take about 90 seconds... but my burst-rates are more than your total BW). It's from a Tier 2 CLEC but the upstream is L3 and Qwest and both Netflix and Youtube work a treat in HD & 4K.
"Completely trivial" != zero, and it's not Verizon's responsibility to subsidize Netflix's business model.
Apparently it's not enough for them to get free bandwidth, when the entire history of settlement free peering has relied on balanced traffic ratios. No, on top of that they want someone else to pay for their rack space, electricity, physical security, and all the other overhead that comes with running a modern data center. Essentially Netflix just wants to take your money and send it directly to Hollywood (after taking a cut for themselves of course) without any of that pesky overhead otherwise known as "running a business".
Well hang on a minute - if I have to put a rack in and provide a standard amount of power to it, that's going to cost me... let's say a few thousand a year at each location (for a large provider, let's go with around 8 locations). So let's take $17,000/year per rack * 8 DCs = $136,000/year as the cost for a major player to run 8 racks of caching servers in existing data-centres around the country.
Even with transit bandwidth being as cheap as it is (in major centres at least), for a 10gig link I'm going to be paying (as an ISP)... probably a few thousand a month, but I'm at their mercy when either end of that link (whether their fault or not) inevitably becomes congested. So taking HE.net's advertised pricing of $0.46/mbit * 10gigs = $4,600/month * 8 locations = $36,800 * 12 months = $441,600.
But wait! We're talking about settlement-free peering! Well, with the peering I'm paying for optics and rack-space anyway - though not as much as I'm paying to run a caching box, I'll grant you. The question then is: private or public? Using Coresite public peering as an example we're looking at a mere $6,000 a year for a 10gig link * 8 locations, so only $48,000 or so, or about 1/3 as much as a rack of caching servers. For private peering... who knows what Verizon's (or Comcasts or anyone elses) costs are.
While the figures are merely an educated guess based on prices I've been able to find in Chicago in public facilities which may bear absolutely no resemblance to Verizon/Comcast/etc costs for the same, so far, running a rack full of caching servers is better than transit, but peering is better than running a rack full of servers.
But that then begs another question - the customer getting what they're paying for. If it weren't for all the anti-class-action clauses in ISP contracts, they'd probably be getting sued a lot, and the TV would be full of commercials for lawyers suing ISPs instead of (or more likely, as well as, pharmaceutical companies)... but generally speaking, most ISP customers in the US are getting ripped off big-time. I myself just got a 50mb cable connection (I was on 100mb fiber, but I moved), and I can get 50 (even up to 70mb) on speedtest.net but Youtube won't give me 1080p or 4k on anything (it doesn't stutter or buffer, it simply doesn't offer me the option, even on videos I know to have it - the best I can get is 720p), and Netflix seems to only run at SD. So I'm marginally annoyed about that.
And if it's such a small cost to upgrade the speeds of the peering links, why aren't they being upgraded? And if peering can no longer be settlement-free due to the asymmetrical nature of the traffic, why can't the ISP combine the two most cost-effective options both to improve the customer experience AND hopefully alleviate some of the asymmetry.
Is $136,000-ish a year really going to hurt that much? That's only... what... a couple of thousand extra customers that will need to be signed up to break even on that investment (assuming about $70/mo/sub revenue less expenses). Plus, with caching servers, I could even use that to attract more customers because sites like this one will have people saying things like "choose this ISP if you want to watch Netflix because caching" (probably... that may also not happen, who knows)...
In the grand scheme of things, it's barely a drop in the ocean of expe
Whenever I've gone to a store looking for a phone, I make it a point to avoid the iSection because I don't want to get any on me.
Last time (less than 3 weeks ago, as it happens) a $CARRIER employee asked "Why not get an iPhone?" when I was clearly looking at Androids, I told him I couldn't take him seriously and waited for another employee who would sell me what I wanted.
So no, I don't believe it was a mistake for a large percentage of people - the customers were probably (in their minds) "graduating" to the top-of-the-line flagship carrier device after having had a feature phone, a low-end smartphone (probably Android) and then a higher end smartphone (probably a better Android) every 2-ish years for the last few contract terms... the kind of people with not a great deal of money to spend so they kind of have to prove to themselves that yes, they can handle having a $700 smartphone (but who still buy the el-cheapo sub-$500 laptops at Wal-Mart and load them up with more malware than I can shake a USB-stick at).
Were they **really** satisfied? Well, chances are the marketing survey was taken while the phone was new, so, in comparison to their old phone, yes, it's going to score 4 or 5 stars. Ask the same people half-way in to their contracts and the results are probably not nearly as pleasing.
Depending on the ISP(s), this type of behaviour might be prohibited by the ISP(s) that are providing services at each of the complexes, as would charging individual tenants for the access and numerous other activities.
Typically even in a dedicated line most providers are OK to let you provide access on a straightforward "as-is as part of the rent" basis but they might interpret such meddling and the additional revenue generated as charging for a service that isn't technically theirs to charge for; so the company in question might want to check the T&C in each of their upstream contracts (eg does the contract permit reselling of bandwidth) before their upstream provider(s) lawyer up and sue for breach of contract.
If, on the other hand, the company has assets like it's own ASN/IP space/FCC registration (and purchases from a non-competing provider such as like Level3 with permission to resell the bandwidth in the contract), then this whole thing might be a different story.
(This information is based on my own experiences running dedicated fiber lines from other some large ISPs and provide access to similar complexes in college towns in a few different states and discussions with various providers in each state).
Yes.
Lets put this all into perspective. Transit is priced about $0.45 retail, and costs even less. But lets give Comcast a bit of wiggle room and assume that on average, bandwidth costs $0.50/mbit. This is bandwidth that can reach anywhere in the world. Comcast oversubscribes about 20:1, which means that bandwidth is really about $0.025/mbit per customer. Comcast then turns around and charges $100/month for 100mbit, which only costs them about $2.5 in bandwidth. Then Comcast complains that Netflix is using too much of this bandwidth, because Comcast's network can't handle providing a measly 5mbit/s to many of their customers at the same time.
If you buy from the right companies in the right quantities in the right markets, yes, that price is accurate. Outside of those markets (smaller towns) with smaller quantities (n*1gb instead of n*10gb or n*40gb) leads to a higher price as averaged across all service areas (plus the cost of the middle-mile). Although in Comcast's case, I doubt the overall rate per megabit would be significantly higher.
Moreover, that bandwidth does not reach anywhere in the world - transit can cost per route kilometer or per megabit (or some combination of both), so they might buy a connection to London which costs say $3/mb but a connection to Hong Kong which costs say $4/mb, and if the destination is in (for example) Singapore, then the traffic is likely to be sent via London unless packets are being dropped, even though sending it via London might be less efficient from a distance perspective. Or in the case of something a bit more domestic, this might mean configuring things such that all Netflix-originated traffic may only come in via the cheaper Cogent links even though the more expensive L3 links might have spare capacity on them - to the detriment of the Netflix service and Netflix customers experience because as far as Comcast is concerned, who gives a crap about the customer.
As far as the costs are concerned, using the same math ($0.50/mb) and contention ratio (20:1) you come up with a cost/mb of $0.025/customer/mb (personally I suspect their contention ratios are WAY higher than that based on the fact that their links are allegedly permanently oversubscribed, but let's roll with it) but strictly speaking, this isn't exactly true - it's not compression - you're not going to magically be able to transfer 6TB/month of data over a link that can in fact only do 300GB/month (I use 300GB as a nice round figure for data transfer per megabit per month - it's a little lower than the theoretical maximum but accounts for network overheads).
Really, the COST doesn't change, just the actual amount of available megabits to the outside world available to each customer across the customer base does, and this is one of the reasons for data caps being implemented (so it would be better to count a per GB cost than a per mbit cost since we're looking at the volume of traffic, not the cost per subscriber mbit as this is only ever going to be measured by Comcast up to their nearest PoP/node/DC). Using a 300GB cap or 1mbit as an approximation, it could be suggested that out of their $100 Internet bill, the average subscriber is paying a grand total of $0.50 for bandwidth (other parts of the infrastructure, CAPEX/OPEX, overheads etc notwithstanding, of course). Which is pretty pathetic, really, and my humble opinion is that now that Netflix and Comcast are directly connected, traffic between those networks should no longer contribute to any data cap that may be in place.
The problem is in part that even though their last mile should theoretically handle a bunch of 5, 10 or even 50mb users all using the network at the same time, and if measured to the nearest Comcast server there's no reason any given subscriber shouldn't get most or all of their subscribed speed most or all of the time... but since we're trying to access the WORLD wide web, that's not how the customer sees it, and as the middle and/or first miles are less able to handle all those subscriber
+1
Kind of like NZ.
Is it worth noting that it's something like $2,000 per premises passed? Even if it were only $700, this difference does change the economics substantially... if you're the 5th person on a block and the first 4 don't also subscribe to the service, they've effectively shelled out $10,000 (or $3,500) to supply one customer rather than the figures often quoted. Of course, if all 5 premises subscribe, the economics are favourable.
Which is why open utility/municipal/government last miles make sense: since the bulk of the cost comes from the roll-out, you've got 5 premises who will be subscribing - even if they are subscribing to 5 different ISPs - and the return on investment is being made on 5 * $2,000 rather than 1 * $10,000 (or 5 * $700 rather than 1 * $3,500 if you prefer) meaning everybody's bills will be lower because the costs can be spread out across more subscribers.
We've heard this argument 1,000 times before but, if you'll excuse the expression, it doesn't hold any water.
Having been to most of Europe and most of Asia, population density in cities on those continents is sufficiently similar to cities in the US so there is no excuse for Internet access in any American city to be any worse than it is in any other city on any other continent, and there's plenty of dark fiber around, so with routing equipment being as cheap as it is now (even at the 10gbit/s level) it's not a question of backbone connectivity either.
Rural connectivity - that is, Internet access to farms or sparsely populated areas - is certainly an issue to be tackled, but there's no reason even small-ish cities can't be provided with decent access.
But we can't compare apples to oranges, either. You can't compare a 100/100 pipe in central Stockholm to a 0.75/0.25 pipe in the middle of Montana. After all, you might be just as likely to get gigabit Internet in rural SK/Japan/Sweden as you are in rural USA unless it's regulated that the ISP or infrastructure provider has to do it even at a loss.
What *IS* different about all those countries though is that the last mile is regulated to be open and shared among competing providers, so they keep their customers by providing better services and faster speeds rather than by protecting their monopolies or suing the competition out of existence and this makes for a fairly significant difference in availability of services.
Amazon can make a profit on 2c/GB because of 1. How much traffic it buys overall and 2. The locations of it's data centres - they're usually in or near fairly major places with IXs to plug in to and cheap transit (and if it isn't cheap at first, it becomes cheap because of #1).
The same is not necessarily true for ISPs. As much as I too dislike Comcast and other such providers and what Comcast in particular is doing, the cost of delivering bandwidth where I am at the moment (Southern Illinois) is not necessarily the same as the cost for delivering bandwidth in, say, Chicago, where I could probably get that cheap $0.46/mbit transit from Hurricane Electric if I needed it there and/or had dark fiber from there to here, but, based on my costs, my per-GB rate down here breaks down to about $0.035 before overheads/taxes so by basic rules of accounting I need to charge at least $0.10/GB to be profitable (which, mind you, is still more reasonable than the overage rates from the likes of Comcast and Mediacom which are $10/50GB or $0.20/GB).
Plus an infrastructure fee - which Amazon doesn't need to charge (again, in part, due to volume).
To put it in perspective, Frontier only has a 2gbit/s pipe to their nearest CO (as of earlier this year), so I'd imagine their costs aren't a whole lot cheaper than mine here (of course, I could be wrong) but in some of their other markets it may be far cheaper, so, IF my services were nationwide (or at least, significantly more widespread as Comcast/Frontier/Mediacom) I'd be taking an average cost of delivery and basing my retail prices on that.
Of course, being that I'm hopefully not an idiot, I'd definitely take a Netflix box in my network if they wanted to give me one. I'm not too concerned about it just yet though - our connections are only 50-60% used during peak hours at the moment.
**sorry for rambling**
Only 2.6 seconds? I've used Internet services on more than one continent that have had things so badly configured my ping times - even to servers like 8.8.8.8 - were more than a minute. Even recently in the US I was using something that gave me pings of 13 or 14 seconds to Google or OpenDNS or something.
We bill them as completely separate entities.
For example, on our most widely used product in India there's a fairly nominal charge of Rs199 plus a flat charge of Rs10 per GB, in the US for a similar product it would (usually) be $20 plus $0.15 per GB. 30GB minimum purchases apply but subscribers can purchase more usage half-way through the month if they want to.
And there's no speed limit on usage-based plans (usually the subscribers port speed is 100mbits, but in some areas we're going to a gig) - I figure when it comes to speed and usage, we should limit either one or the other, not both. At least not artificially.
What if you're like millions of people both in this country (I assume you refer to the US) and others who are tax-payers but not citizens? Like myself. I pay tax in more than one country that I am not a citizen of... so unless you're suggesting non-citizens can get out of paying tax...
Ergo, really, taxpayer really does seem like a more appropriate term when it boils down to it, because citizenship comes with certain rights that aren't afforded to non-citizens (so, you're right about the "weight" thing) but when it comes to the funding of government through taxes, it makes no difference whether the individual is a citizen or not because if all else is equal, the tax burdens of either are equal and their taxes are spent the same way on the same stuff.
Interesting you should say that... this is exactly how we bill some of our plans. So I guess I agree with you.
There are lots of such routers now. Ubiquiti EdgeMax Lite ($99) springs to mind but there are a bunch of others as well.
Granted, configuration may not be as straightforward as your average home router and you'll end up with 4 devices for 2 connections (1 router, 1 wifi, 2 modems) but it has an half-way decent GUI and a wizard so it shouldn't be out of the question for someone who is at least partially competent to set it up for dual-wan.
Also, I think the Zyxel modems that Frontier give out have dual-ADSL ports in them but they're not very good IMO. I don't think they're terribly expensive though.
I don't know where you are but "several thousand" sounds like AT&T/Verizon money to me - I've been quoted over $5k for 100mb both here (Southern IL) and in other states while the local CLEC charges a mere $1,000/month for 100mbit symmetrical / dedicated fiber (not 95th percentile) with all the usual things like a /29 and 4 hour MTTR - (it's about $750 for 50mbits and $400 for 10mbit/s I think)... once the packets hit St Louis we're on Level 3 and Qwest which means most stuff people are accessing is pretty speedy and there doesn't seem to be any congestion.
And unlike AT&T and Verizon, the more you buy, the cheaper $/mb and they aren't using crappy bottom-end Cisco routers.
$10/mbit works out at a cost of $0.035-ish per GB of data and while I'm not talking middle of nowhere, I'm not talking about a large market either.
I think your numbers are a bit off, both as far as the price of hardware vs ROI and as far as the "huge profits" are concerned - aren't Comcast and Verizon supposedly making 40-90% profit on multiple Billion dollar revenues? (Granted, for these companies this may be an overall company margin including all business units, not necessarily the margin on HSI).
In any case, my own companies are much smaller than them (so higher cost per customer for equipment and more expensive bandwidth because I'm not buying nearly as much as they are so I don't get quite the same volume discounts) and if it weren't for our expansion projects, we'd be doing pretty tidily.
And assuming people aren't dicks about bandwidth usage (eg hogging all of their bandwidth all of the time), we can even make residential services a little bit profitable without cross-subsidizing between plans (eg higher tiers aren't subsidizing lower tiers and vice versa). In the US, we'd be looking at something around 80% profit before equipment & overheads, and with a PON setup working out to under $1,000 per port (assuming brand-name gear, 10gb optics in the back and GPON for the customer-facing ports), our RoI really isn't as long as one might expect.
Population density per square mile as averaged over the entire country is really not relevant. Someone living in the sticks is going to have limited connectivity options whether they're in the UK, USA or most other countries (even our favourite poster-children like S Korea and Japan).
A town/city with 2k or 20k or 200k or 2m people is a city with 2k or 20k or 200k or 2m people in either country and, geological differences and labour costs notwithstanding, //should// cost approximately the same amount of money to service (and thus, retail prices should be pretty similar).
I put to you that if 2 last-mile projects were simultaneously undertaken to provide 1,000 customers in 10 a square mile area, but one was allowed a monopoly on it's last mile and one was not and had to be open/unbundled and had 5 operators competing for those customers as detailed by cid=46867081, the latter would have far cheaper end-user pricing, irrespective of which country/ies we were talking about.
Really? $0.25/mbit? In what volume? In which market? By who?
In many European countries, food & housing is subsidized. 0.80 euros a meal (buffet style) at the University cafeteria and about 350 euros a month to live in student housing (1 bedroom apartment).
And this was in the mid-2000s, in Finland - one of the more expensive countries to live in - which meant that the Internet connection (which was included as part of the rent along with other utilities) was better even then, than I could get now in 2014 in the US (if I were a regular Internet subscriber, that is).
And car? WTF would you need a car for in most European cities? Students in most European cities would not consider a car to be a necessary item (nor would many adults) and since public transport costed about 40 or 50 euros a month for a pass that got me on any form of Helsinki public transport (bus, local train, tram) which ran several times an hour, a car simply wouldn't have been worth it.
And having also lived elsewhere in Europe, I seem to recall that situation being pretty similar in many other parts of the continent too.
I got asked by a CBP officer on one of the occasions I came in to the US last year if I paid bribes in India (I think it was at ORD).
I answered him honestly (I'm neither American nor Indian) to which he replied that I was brave for doing business there and sent me on my way.
Not just in Russia, that's for sure.
I dunno, have you ever used the Internet in Bulgaria? It was pretty good last time I was there.
Albania on the other hand...
In many countries (especially in Europe) modems/cable boxes/etc are included in the price of the service - I don't seem to recall being charged extra for ISP-supplied equipment in France, Japan, Finland, Georgia, Ukraine or New Zealand**, but either perhaps I didn't notice or it's simply built in to the price over the term of the contract (that is, it's an implied overhead so it's accounted for by locking you in for 12-24 months)... but typically you're on a term contract in the US as well so there's no difference there anyway.
And I seem to remember Free & Neuf boxes being valued at something in the vicinity of 200 euros at one time as well, so they're no cheaper.
**May also vary by ISP but may also be entirely wrong about the whole situation in some or all countries - correct me if I'm wrong.
Especially when "tips", as it turns out, are supposed to legally make up the actual hourly rate paid to the server to meet minimum wage, for example, the server's hourly pay is $2.50, it is expected they will make at least $6 in tips per hour causing their total hourly rate to equal the $8.50 minimum wage (numbers loosely based on Illinois, I don't know specific wage rates for other states)... so if customers stiff the server on their tips, the server gets screwed.
As a non-American though, I agree with you: tips (and prices being advertised without taxes, for that matter) are bloody annoying - I used to get caught out all the time when I first came to the US. I'd hate to be in the service industry here.