Wednesday, August 15, 2007

BBC's iPlayer as the posterchild for net neutrality

It's interesting to see how in the UK some of the lesser ISP's (Tiscali and their lot) have complained in the press about the public broadcasting behemoth BBC and its iPlayer. iPlayer is the BBC's attempt to copy the Dutch public broadcasters success of "Uitzending Gemist" (=missed my program!). It does this by using a Peer to Peer program Kontiki. The ISP's claim that this is at their expense. IPdev-Blog and Telebusilis have analyzed this in some detail.

Jeremy Penston of IPdev has analyzed very well why ISP's won't invest themselves in new networks and network expansions. The process in short is one of mutually assured destruction. If two companies build the same network, they create an over supply of network connections and bandwidth in the market. They will end up in a price war where neither can bail out and both will go bankrupt in the end. (even if one of them wins the first round, the loosing network can be revived from bankruptcy at marginal cost and start the second price war) The solution seems to be a regional or national public infrastructure. I agree with his ideas and hope to publish a paper along those lines soon.

However both Telebusilis and IPdev argue that the content creators should finance in one way or the other the build-out of extra capacity in the network. They argue it is not fair for the BBC to come up with a new service that taxes the networks of ISP's in Britain (up to 67pence per hour of viewing). I couldn't disagree more with them. I think it's only the ISP and it's customers that should do so. It's the end-user that creates the costs and it's there the costs should lie.

We live in great times, on a daily basis people al around the net invent new high bandwidth services to use over the internet. I'm watching my 3 day old cousin in a hospital on a high def webcam. You can watch live concerts at Fabchannel. People dress up in Second Life. In Twente security companies watch their customers premises using dedicated light paths. Every tv-channel and production company is looking into the on-demand opportunity. These new ideas have ever higher bandwidth demands.

In order to minimize the costs for content producers there are several strategies. Bill Norton of Equinix has made a very good analysis of the costs of video distribution over the internet. His analysis shows that using a Peer to Peer model (like the iPlayer) is the most cost effective version for the content provider. Or as Cringely paraphrases it:

Norton's analysis, which appears to me to be well thought-out, concludes that P2P is vastly cheaper than any of the other approaches. He concludes that distributing a 1.5 gigabyte movie over the Internet in high volume will cost $0.20 using the current transit model (a single huge distribution server), cost $0.24 using an edge-caching CDN like Akamai, cost $0.17 with a homemade CDN like I used last season to distribute NerdTV, or cost $0.0018 to distribute using P2P. That makes P2P 35 times cheaper than any of the alternate approaches. And (...) Norton further makes the point that none of these distribution models does anything to soften the blow on the ISP. CDNs in particular cost more -- that more being revenue to the CDN -- yet do nothing for the ISP.

Well the BBC could also do the calculations and came up with the advised solution. Which might actually be a better solution for ISP's as well mind you. This not often mentioned, but a well designed p2p-protocol keeps local traffic local. So if your neighbour wants to watch a movie that you so happen to have on your pc in an ideal world he would not need to burden the backhaul-links from your town to the main switch office, but keep everything local. This relieves the network of the ISP from heavy backhaul traffic. Just imagine if an entire town would be streaming from the servers of the BBC. At 1 megabit a town with 10.000 parallel streams would be hitting 10Gbit/s on the backhaul. This way the ISP can save on its backhaul and also on its interconnects with eg the BBC. (How perfect the world of p2p protocols is can be seen at IPdev here and here)

So why shouldn't the bandwidth hogs be paying for their bandwidth? The BBC has enough money and they do pay for satelite capacity, so why should they get away for free. Well, the BBC isn't they only one designing high bandwidth services, as said it's everybody. All those new services mentioned contribute to the networks creaking under heavy loads. Remote security cams, baby cams, people with no First Life. All of them break the network. Even normal surfing the web helps. The question about who pays than quickly becomes a question of who can we extort money most easy from. Well auntie Beeb is old and wealthy, so it might be easy to beat her up for her pension. It's much harder for a UK ISP to do the same from a Dutch hospital, or security company or a Japanese public broadcaster, though they might contribute as much to the demise of individual links as the BBC does. (Think of it as cars on the road, all the cars contribute to congestion, foreign and domestic, business or pleasure). So what you get is that the costs are disproportionally allocated to those companies that are easiest taxed.

Another reason against using a taxation of content providers is that the revenue stream will be so attractive to improve the competitiveness of the ISP that there is no reason to assume the money will go into network upgrades. They might just as well go in more advertising or lowering prices. Even better, there is no reason to expect the taxation to cease once the network has increased its capacity. Like so many taxes they tend to linger long after they´ve done their job. For the economist, it's kind of like a terminating monopoly and will require equal amounts of regulation.

A third reason is that imposing a `Save the ISP`-tax is detrimental to innovation. Think of it, would you want to father the new Skype if the bandwidth tax bill ends up on your doorstep? Ofcourse not. That would be ridiculous.

By now people will be confused. It must be expensive to get a new network that can handle this amount of traffic they think. But again they are wrong. You can get a nationwide fiber to the home network for roughly 35 euro per house per month (or an investment of between 1000 and 2000 euro). For most countries that is signficantly less than their investments in roads and it is equal to what it would cost now to build an electricity network from scratch. Yes there are upfront costs, but it would last 50 years, allow for all kinds of innovations etc. If the market doesn't provide this, you have a market imperfection that might require limited government intervention in the civil engineering part of the physical network, if the benefits outweigh the costs eg. Stokab in Sweden. But there are billionaires around willing to do cherry picking in FTTH networks (Dik Wessels with Reggefiber). And there are even smart incumbents upgrading there networks to VDSL2 (KPN, Deutsche Telekom) or FTTH (Verizon) and new entrants (Free). Though we are still a bit away from universal 1 gigabit home connections for 35 a month.

35 euro per month buys you the fiber network (less if we fix it partially with government money). Interestingly it doesn't matter whether you use this at 1 mbit/s or 100 mbit/s or even a gigabit. It all costs exactly the same. Different speeds of your ADSL line eg 8mbit or 1mbit are only a way of price differentiation, but have nothing to do with sending more bits over the network being more expensive. It doesn't get you the traffic yet. International and interregional traffic costs money. The way this is dealt with in many countries is with monthly traffic caps eg of 40 Gigabyte and if you use more you pay more or there is an acceptable use policy. The way this could be fixed in the future is that you have a gigabit line to your house and a terabyte of traffic per month of interregional/international traffic (local traffic is free). If you go over, you pay more.

Now we arrive at the problem with high bandwidth applications like P2P applications or babycams. The way Joost and BBC's iPlayer work is that they exchange traffic even when users use it. Users actually have no way of knowing or limiting the amount of traffic it uses. With a babycam you could calculate it, but it's not intuitive. This should be fixed. A user should know how much costs they are incurring by using innovative appplicatons. They can then limit their usage according to their needs. It will also push ISP's to increase the monthly traffic cap to offer their customers more than the competitor. ISP's can now extract the money from their customers based on the amount of bits and not on the type of application or which granny to beat up. If a customer wants to use more they pay the ISP and they get the bits, regardless what they use them for.

Alright this seems too easy. Networks get paid for by the customer and it seems like content providers are getting a free ride on the network innovation train. The content providers have all this income from advertising and they should share... shouldn't they? There are several arguments against this. First of all, it's highly questionable if there really is so much money in advertising. The total turnover of the Dutch advertising industry is 6 billion and this supports Ten TV-channels, around 10 national newspapers and a couple of hundred magazines, thousands of websites etc. Some of it doesn't even support content, like billboards and classifieds systems like Monsterboard. (In comparison the mobile telecoms sector makes 6 billion a year too with 4 networks) Secondly efficiency in distribution leaves room for innovation elsewhere. Just like containers revolutionized shipping and realized China's position factory of the world. So too will new networks and p2p decrease transaction costs and revolutionize the delivery of content. This will lead to globalisation of the content market and the infrastructure will lead enable all kinds of innovations from babycams to immersive content. If there are excess profits to be made in the content market by advertising and pay-per-view models, there will be new entrants into the market, the breadth and hopefully the quality of the content will go up. This will redistribute the wealth in the market to such an extent that the big advantage of content owners over ISP's that some see will disappear. Efficient markets hate long term excessive profits for an entire industry. Though one compnay may prosper because of enormous economies of scale and network effects.

Therefore the conclusion is:
New applications will demand more and more bandwidth. Their combined usage will compound to the problem. This will push ISP's to deliver more bandwidth and traffic. Users will be paying for this one way or another. If the market doesn't provide for bandwidth, government should. ISP's taxing those who design applications that use high bandwidth is not a solution, it would be a disaster. We need innovation both in content as in applications and services In order to relieve backhaul local traffic should stay local and local interconnection should be possible between ISP's and private networks, see NDIX for a great example (yes I once worked there).


Jeremy Penston said...

Hi Raindeer,

You should write more often. There are some good points in there - I particularly like the innovation point and the one on taxation not going away.

That said, I do have to disagree with the implication that users see value in the extra megabytes they get and that these can be somehow billed. There is some value still left in megabits per second because that is something we understand (fast vs slow).

By using more megabytes is such a random and unpredictable thing that I just don't see how humans can make acurate judgments about how much they are using.

The counter to this is whether we understand kilowatt hours of electrictity which I for one don't either.

But then if I brought home a new domestic appliance that ended up doubling my electricity bill I would have an issue - with both the manufacturer and the electricity company. It would leave a sour taste all round.

No, usage based billing works only when the the impact of any one transaction is small and if not known, known not to be a problem. We may get there when there is fibre everywhere and a much broader scope of applications like you outline in the piece, but we have some steps to take before we are in that situation - at least in the UK ;-)



Raindeer said...

Hi Jeremy,

Thanks for the positive words. Do have a look at the archive and I will post some more in the future :-).

I agree with you on the pricing side, the cost should be negligible, but with 1mbit (roughly 200 Gbyte effectively, taking one third off the top) selling wholesale at less than 10 euro per month. I think we're pretty far in that land already. I know that in the UK, the Wholesale Broadband Access people have some weird charges to suffer through 67p?? WBA bandwidth is more expensive than transit, or to put it differently. It costs at least 5 times less to get the bits from Tokyo to London than from London to Edinburgh.

The regulator should look into putting some dynamics into the pricing of backhaul links.The amount of money for backhaul can be seen as a fixed amount per month. There are onlyfixed costs, wages, depreciation etc., but no variable costs that change depending on the season. With current traffic growth being at between 50% and a 100% we would see a yearly decrease in traffic/bandwidth charged per Mbit/s/month of 33%-50%. With the right depreciation scheme and sound investment you can include the upgrades of backhaul links from 1Gbit/s all the way to multiple 10Gbit/s and still see the prices per mbit/s/month drop. Best thing almost would be to do this on a monthly or quarterly basis. But that's up to the regulator



Esky said...

I don't for one second think ISPs should be taxing the BBC, and generally agree with your points. I did just want to correct one thing you posted though, as I think many people misunderstand this. For ISPs that aren't utilising LLU and who want to provide nationally available products, you are pretty much tied to a BT IPstream based product. Using IPstream means a couple of things:

- Your ADSL connection to your ISP is an ATM tunnel (That's what the A stands for in PPPoA). As such, no local aggregation of content is possible. If you share content with your neighbour over, BT send that data back to the ISP where the customers connection is terminated. The ISP then routes it back out to the other subscriber, back over the BT network. In short, the ISP has to pay for that data twice! 21CN should improve on that because it's a very different Ethernet architecture, but until we have that in place, this is going to be a really tough problem to solve.

- The cost of IPStream is between £180 - £200 per Mbps, depending on how you do the maths. Internet bandwidth itself is about £10 per Mbps. I don't think people realise just how much ISPs have to pay - One customer downloading at 7Mb/s for a month easily costs £1500!

In any case, a good article, and nice to see this discussion happening.

Ian Wild
PlusNet Product Team

Raindeer said...


I can only say: YIKES!!! or OUCH!. Those are serious amounts of money. I'm guessing Ofcom isn't kind enough to review the traffic prices on a quarterly basis. So though bandwidth demand is growing steadily 50%-100% on a per user basis, wholesaleprices aren't dropping and therefore the consumer price should be increasing. (Though somebody using 7mbit/s continuously for a month would use 2Terabyte and that is quite a heavy user, but you're right)

This does bring the following question to my mind. Under what conditions is BT connecting it's users? BT is now split up in a wholesale Openreach part that only doest the network and the services part that is an ISP. Now does BT only work under ULL-conditions and does that mean it is not subject to these extremely high backhaul costs? If so, that's tough cookies for the IPstream users. There is no way they can compete. Better said, even ULL providers will be out of luck, since BT can just provide everything everywhere under favorable conditions.

Esky said...

Well, it's interesting you should ask about BT. BT retail use a product called Central Plus, which economically is actually much closer to IPStream than it is to LLU. Bt Wholesale (Who sell Broadband capacity) and BT Retail (Who buy it) are regulated and must operate independently and profitably as individual business units.

Basically, right now ISPs like BT are able to off-set their light users with their heavier users and as such average usage costs in. That is set to change though as the lighter users discover Youtube, the iPlayer and 4OD etc. The question is all about when that happens and whether the next product generation from BT (Wholesale Broadband Connect) will price competitively against LLU pricing. For the moment though I can certainly imagine things are going to get squeezed for many, including BT retail.

A friend and I are writing a blog post to cover the economics of this a bit more. It will appear here later this week:

Drop me an email - iwild at plus dot net if you would like to discuss any of this further.


Anonymous said...

Here’s a Broadband Video that will show you how to check availability by postcode, how perform a broadband speed test and where to find broadband forums to answer your questions. There are also offers for BT Broadband.

Here is a broadband beginners guide and below are common broad band questions;
What is Broadband?
Types of Connections?
Connection speeds and download limits
How to compare ISP’s