BBC’s iPlayer nukes “all you can eat” ISP business model
The UK’s largest broadcaster finally launched its online video streaming and download service on Christmas Day. Plusnet, a small ISP owned by BT, has provided a preliminary analysis of the traffic and the results should send shivers down the spine of any ISP currently offering an unlimited “all-you-eat” service.
The iPlayer service is basically a 7-day catch-up service which enables people who missed and didn’t record a broadcast to watch the programme at their leisure on a PC connected to the internet. The iPlayer differs from any other internet-based video service in certain key respects:
- It is funded by the £135.50 annual licence fee which pays for the majority of BBC activities. The BBC collected 25.1m licence fees in 2006/7. No advertising is required for the iPlayer business model to work.
- It is heavily promoted on the BBC broadcast TV channels. The BBC had a 42.6% share of overall UK viewing in 2006/7 and therefore a lot of people already know about the existence of the iPlayer after one month of launch.
- It is a high quality service and is designed for watching whole programmes rather than consumption of small vignettes. This is sharp contrast to the current #1 streaming site, YouTube.
A massive rise in costs
The key outputs from the Plusnet data is that in January:
- more customers are streaming;
- streamers are using more; and most importantly
- peak usage is being pushed up
This equates for Plusnet to streaming cost increasing in total to £51.7k/month from £17.2k, or an increase of 18.3p/user from 6.1p/user. This is a 200% cost increase in just the first MONTH of the service.
If we assume that the Plusnet base of 282k customers is a representative sample of the whole UK internet universe than we can draw some interesting conclusions about the overall impact of the iPlayer on the UK internet. On the whole UK IPstream base of 8.5m the introduction of the iPlayer would equate to an increase in costs to £1.5m in January from 500k.
Despite access unbundling, ‘middle mile’ costs remain a key bottleneck
IPstream is a wholesale product from BT, with BT being being responsible for the transit of the data from the customer’s home to an interconnect point of the ISP’s choice. The ISP pays for bandwidth capacity at the point of interconnect. BT Retail acts like an external ISP in the structurally separated model. The overall effect of the iPlayer for the BT’s IPstream-based customers is roughly neutral, with the increase in revenues at wholesale (external base of 4.2m customers) being offset by the increase in costs at BT Retail (total base of 4.2m customers). Of course, this assumes no bandwidth overages at BT Retail, which probably is not the case as both BT and Plusnet have bandwidth caps.
In effect, incremental cost for ISPs using the IPstream product is determined by ordering extra BT IPstream pipes which come in 155-meg bit size chunks. The option for the ISP is either to allow a degradation in performance or order more capacity.
Time to buy more pipes
We tested the bandwidth profile using Wireshark watching a 59mins documentary celebrating the 50 year anniversary of Sputnik with both streaming and P2P. The streaming traffic is easy to analyse as it comes through on port 1935, which is the port used by Flash for streaming. Basically a jitter-free screening ran on average at around 0.5Mbit/sec. Using the 155-meg ordering slice this means only around 300 people need to be watching the iPlayer at the same time (peak = 8pm-10pm) to fill a pipe.
Seeing that IPstream customers are aggregated across the UK to a single point, a lot of ISPs will be thinking of the need to order extra capacity. The BBC also offers a P2P download which is of higher quality than the streaming. We managed to download the 500Mb file in just over 20 minutes at an average speed of 3.5Mbit/sec. The total traffic (including overhead) for the streaming was 231MB and for the P2P delivery was 544Mb.
Full unbundling still leaves ISPs at the mercy of backhaul costs
The story for facility-based LLU players, which account for another 3.7m UK broadband customers, is slightly different as it depends completely on network design and distribution of the base across the exchanges. Telco 2.0 market intelligence says that some unbundlers have ordered 1-gig links for the backhaul and should be unaffected least in the short term. However, some unbundlers have only ordered 100-meg links and could be in deep trouble with peak hour people really noticing the difference in experience. The only real option for these unbundlers is to order extra capacity on their backhaul links which could be extremely expensive. The average speed for someone just browsing and doing emails is quite low compared to someone sat back watching videos stream.
Cable companies understand sending telly over wires
The story for Virgin Media, which is the main UK cable operator with 3.3m broadband subscribers, is again is dependent on network design. This time it depends upon the load on the UBR within the network segment. Virgin Media have a special angle to this as the iPlayer will be coming to their Video-on-Demand service in the spring, and therefore we assume this will take a lot of load off their IP network. The Virgin VoD service runs on dedicated bandwidth within their network and allows for the content to be watched on TV rather than PC. A big bonus for the Virgin Media subscribers.
Modelling the cost impact
For both cable and LLU players the cost profile is radically different to IPstream players, and it is not a trivial task to calculate the impact. However, we can extrapolate the Plusnet traffic figures to note the effect in volumes of data.
We have modelled four scenarios: usage the same as in Jan 2008 (i.e. an average of 19min/month/user) rising to 1 hour/month, 1 hour/week and 1 hour/day. These would give an increase in cost of £1,035k/month, £3,243k/month, £14,053k/month and £98,638k/month respectively for the IPstream industry, only based upon Plusnet cost assumptions. Of course this is assuming the IPstream base stays the same (and they don’t just all go bust straight away!). Across the whole of the UK ISP industry, the increase in traffic (Gb/month) is 1,166, 3,655, 15,837 and 111,161 respectively. That’s a lot of data.
The obvious conclusion is that ISP pricing will need to be raised and extra capacity will needed to be added. The data reinforces our belief expressed in our recent Broadband Report that “Video will kill the ISP star”. The problem with the current ISP model is it is like an all you can eat buffet, where one in ten customers eats all the food, one in a hundred takes his chair home too, and one in a thousand unscrews all the fixtures and fittings and loads them into a van as well.
A trigger for industry structural change?
An interesting corollary to the increase in costs for the ISPs is that we believe that the iPlayer will actually speed up consolidation across the industry and make the life of smaller ISPs even more difficult than it is today.
Additionally because of the high bandwidth needs of the iPlayer, the long copper lengths in rural England and the lack of cable or LLU competition to the IPstream product, we believe that the iPlayer will increase the digital divide between rural and suburban UK.
The iPlayer also poses an interesting question for the legion of UK small businesses who rely on broadband and yet don’t have a full set of telecommunications skills. What do they do about the employee who wants to eat their lunch at their desk whilst simultaneously watching last nights episode of top soap EastEnders?
Time to stop the game of ‘pass the distribution cost parcel’
The BBC is actually in quite a difficult situation, especially as publicity starts to mount over the coming months with users breaking their bandwidth limits and more or more start to get charged for overages. The UK licence payers expect they paid for both content and distribution when they handed over £133.50.
In 2006/7, the BBC paid £99.7m for distributing its broadcast TV signal, £42.6m for its radio signal and only £8.8m for its online content. This is out of a total of £3.2bn licence fee income. I would suggest that the easiest way for the BBC to escape the iPlayer conundrum is for them to pay an equitable fee to the ISPs for distributing their content and the ISP plan comes with unlimited BBC content, possibly with a small retail mark-up.
The alternative of traffic-shaping your users to death doesn’t seem like a great way of creating high customer satisfaction.
The old media saying sums up the situation quite nicely:
“If content is King, then distribution is King Kong”
[Ed - to participate in the debate on sustainable business models in the telecoms-media-tech space, do come to the Telco 2.0 ‘Executive Brainstorm’ on 16-17 April in London.]
Comments
It's such a waste to deliver the same old content to millions of people (soap operas especially) using unicast technology across the whole country from ISP to user. The problem is that BTWholesale gets paid by the ISP per "bit" of content transmitted, so it is not in BT's interest to develop a Content Delivery Network-style approach inside IPStream or the 21CN-WBC. (Or perhaps it will, but at a tidy price most likely. There have been some discussions on a Content Connect product at Consult21 workgroups)
A CDN or peer-to-peer local traffic turnaround option (at the exchange or metronode) would be much more efficient in the long run. Because the episode of Eastenders you want to download was most likely watched recently by someone living near you. With LLU this is likely to be achievable technically and will happen as soon as it is economic for operators. But with BT's disincentive for IPStream I fear your conclusions regarding the digital divide will be true.
Ofcom designated "Market 1" will see poor speeds, poor quality, low levels of choice and higher prices. What we need is a minimum Universal Service Obligation for Broadband, including in time minimum standards for video transmission. However at Openreach backhaul pricing this will be expensive. More investment needed in backhaul to remoter regions needed!
Posted by: Neil | February 20, 2008 2:15 PM
People may be interested in these posts by Ashley Highfield from the Internet Blog which are relevant to the above article:
http://www.bbc.co.uk/blogs/bbcinternet/2007/11/the_net_at_breaking_point_pt_2.html
http://www.bbc.co.uk/blogs/bbcinternet/2008/01/the_digital_divide.html
Posted by: Nick Reynolds (editor, BBC Internet Blog) | February 20, 2008 2:25 PM
Umm, don't the BBC need to pay for their upstream data too? After all, the new iPlayer as a streaming service hits them just as hard, no?
Posted by: Paul | February 20, 2008 6:03 PM
Yes, that is perfectly true.
The BBC are paying for an upstream content delivery network. The source of the streaming data we looked at came from the Akamai server farm so I can only assume Akamai are the people who the BBC are paying an ever growing bill to.
However, the main point of the article is that cost engineers and planners at ISPs will have estimated average peak hour bandwidth consumption for the various ISP pricing plans. The sudden success of the iPlayer will no doubt mean that some of these estimates will be too low and therefore either profits drop for the ISP or prices rise for the consumer.
For the BBC it will cheaper for P2P downloads rather than streaming. I gather that the current streaming to P2P ratio is 8:1 and the BBC hope this will decline to 4:1 over time. Personally, I doubt this and this the mass market will always prefer streaming to P2P.
Posted by: Keith@Telco2.0 | February 20, 2008 7:08 PM
The BBC's upstream is not as expensive as you might expect. The BBC has for a long time been a strong proponent of peering. They just need to get their pipes to Linx et al. The original driver for this was the fact that the BBC News website is one of the most popular in the UK.
http://support.bbc.co.uk/support/peering/
And of course the BBC has investigated multicast, and is undertaking a trail at the moment:
http://support.bbc.co.uk/multicast/
These small ISP's should be beating down the BBC's door, asking for a multicast, cached version of the iPlayer. The BBC should have reasonable stats on their popular programs, and could multicast them. Coupled with the users software selectively caching programs based on viewing, you start to get somewhere nice.
Of course, BT is well aware of the potential of this problem. Their BT Vision product supports PVR functionality of UK DVB transmissions. This should insulate them from a good chunk of iPlayer use.
Alternatively the small ISP's should be meeting their financiers to talk about LLU. Or at least, switching to an alternative provider to BT.
Posted by: Mark Swanborough | February 21, 2008 12:09 PM
The fact that an ISP has to buy capacity from BT in increments of 155 Mbps is a strong indicator of an ATM-based backhaul. Which is far more expensive than an Ethernet-backhaul. A switch to Ethernet would reduce (but not remove) the problem, especially for anyone who owns their own dark fiber backhaul network
Posted by: Herman Wagter | February 21, 2008 2:58 PM
Ashley Highfield has now responded to some of the points in this article in this blog post:
http://www.bbc.co.uk/blogs/bbcinternet/2008/02/iplayer_figures_and_feedback.html
Posted by: Nick Reynolds (editor, BBC Internet Blog) | February 21, 2008 5:53 PM
Keith@Telco2.0 makes a very good point about the BT Multicast trials.
however, these trial have been going on for a very long time now ,infact it was fully open and no registration was required at the time (way before the unicast IPlayer became public)
and so far the biggest players likes of Virgin Media and BT retail
did not respond to the BBCs call for multicast peering inside their network.
and DO NOT allow the Multicasting protocol on their networks to the end users.
madness on the ISPs part as no extra casts were involved and a total ignorance of what the end users wanted to do, I.E join this BBC Multicast trial, but couldnt.
infact, its a very limited list of smaller ISPs that infact did/do leave this Multicasting capability active all the way to the end users.
its shameful to think that every UK based ISP already has the Multicast capability in all their router and related kit ,all the way to the cable/DSL modem sat on the users desks, but the ISPs have turned it off and refuse to re-activate it.
this available yet unused multicast protocol is madness, given the fact that if they just re-actived it, and allowed the users to take advantage of the protocol, the UK might finally see the many P2P torrent and related apps take up the Multicasting protocol and save VAST amounts of collective bandwidth.
anyone remember the old MBONE experimental Multicasting network, that still exists on the JANET network that uses the likes of the core Virgin Media and BT networks so its clear Multicast can work.
its a shame the options always been there but some numptys inside the UK ISPs dont realise multicasting can save them vast amounts in a very short timeframe.
ask your ISP to re-instate the Multicast option on all the internal and external networks and watch as the UK users and JAVA azureus torrent developers move over to the far better multicast option.
hell, it would eveb be in the UK ISPs interest to just take the open JAVA AZ codebase and help retofit a usable multicast extension into the code in a matter of days and save untold bandwidth in a very short timeframe.
failing that , theres always the option for some coders to take the existing java multicasting Bamboo DHT codebase and the Mtunnel java app and comine the two into the tunneled multicast p2p thats been needed for a very long time now.
see:
http://www.cdt.luth.se/~peppar/progs/mTunnel/
http://bamboo-dht.org/tutorial.html
http://www.samrg.org/p2pm07/
Posted by: david | February 23, 2008 10:53 PM
Also, Marshall Eubanks and friends' Automatic Multicast Tunnelling project have got the Beeb to turn up support on their side (basically it permits participating multicasters and recipients to tunnel multi through a non-multi enabled network), but nobody's using it on the recipient side.
Posted by: Alex | February 25, 2008 4:24 PM
In the ISP world Service Models based around DPI and tiered services have been available for a couple of years now, but ISPs have eschewed these in place of simple solutions and easy marketing claims. IP Multicast technologies have been working in Enterprise networks for at least ten years, but transmission focussed telcos in particular have been slow to move from the old world of point to point circuits to point to cloud models, thus helping to create the current issues.
On the consumer side it seems that the misleading "unlimited" broadband messaging that ISPs have used is coming home to haunt them. Subscribers trusting that what is sold to them actually does what it says on the tin, leaving aside the weasel words on "fair use" policy, have had their expectations set and are now attempting to use what has been sold to them.
A solution can be found but the root of the problem is caused by the ISPs themselves and how they have developed and marketed their services.
Posted by: Huw | February 25, 2008 4:53 PM
All the multicast comments are interesting, but BBC iPlayer is on-demand content.
End users will want to watch whenever they want, pause it, fast forward etc. so I'm not sure how multicast offers any benefit to the ISPs.
Posted by: Neil | February 25, 2008 5:21 PM
Multicast offers the capability to allow intelligent distribution down to delivery points closer to the user eg Exchanges.
So you stream unicast from a stream engine close to the user, but use multicast to populate those stream engines.
The cost to an ISP is usually the bandwidth into BT's IPStream network, or in the backhaul from the LLU exchange. Multicast could save on that cost.
This type of model could work well for a catch up TV service.
Depending on cost, a hybrid model with popular programs being prepositioned via multicast and less popular ones only being retrieved on demand could significantly reduce the costs across the BT centrals or LLU backhaul.
The right mix depends on cost.
Posted by: Huw | February 25, 2008 6:44 PM
"Neil:
All the multicast comments are interesting, but BBC iPlayer is on-demand content.
End users will want to watch whenever they want, pause it, fast forward etc. so I'm not sure how multicast offers any benefit to the ISPs."
well thats the thing Neil, at any given time of the day or night theres a very good chance that more than one single person wants to use that (i assume you mean) so called 'Realtime' on-demand content.
if there is only one person wanting that data stream then its just as good as the current Unitcast model, however if even just twp people are using the proposed Multcast service then theres a saving right there, and as more join, the greater the saving YES?.
as unlikely as it might be, but if its found that a realtime Multicast on-demand service cant scale far better than the Unicast model, then just move the posts a little and make it a Near realtime Multicast on-demand service.
that just means you as the UK end user que your on-demand content and every 5 or 10 minutes, the server end or the P2p peers collect this request and send it.
the end users will just as now with Unicast,spool/buffer the data on their harddrives (temp dir) and still have the ability to do all that you require from the local stored data stream.
and they will also be in a position to send the data packets back out to the new requests coming in for the data as per an extended Multicast P2p model.
see the samrg.org Multicast whiteboard paper above for instance.
Alex: whats this 'Marshall Eubanks and friends' Automatic Multicast Tunnelling project' that you spealk of and how does it work?
you make it sound as thought the BBC have infact setup the Multicasting Tunel at their end that i have been requesting for a very long time now?.
if thats the case, doest that imply i can use any Multicast streaming video app such as VLC or MPC to now join a BBC Multicast tunneled IP today over my Virgin Media IP connection?.
if so ,please feel free to give all the info required and the URL to use etc and start advertising this trail service ASAP to get the word out and put the wastful Unicast to bed ;)
Posted by: david Manchester | February 26, 2008 4:15 AM
How, while your hybrid model would work, i feel its about time any Multicast restrictions are totally removed today.
its wrong that the end users and the developers cant make use of IP Multicast or IP broadcast for that matter just because the ISP want re-activate Multicast etc on the router and related kit.
once it is reactived all the way to the end users and back, you will see the popular (P2P,Video,interactive white boards apps etc)get reto- fitted with this ability and in the process we might start to see some real innovation in the new apps and services that can be provided.
after all your average users dont mind the business making a reasonable profit IF they can see a benefit.
and Multicasting in todays mass bandwidth market is the perfect time to activate it as the compulsary Docsis3 (bronze) Multicast and (silver)Qos Mulicast come online, plus the new BT 21 Multicasting abilitys too.
i really think its wise to reactive Multicasting NOW rather than wait for someone else to do it.
as per the Virgin Media model of wait and see what the others do, means you are going to loose big time instead of capitaliseing on your available installed kit while you can..... ;(
Posted by: david Manchester | February 26, 2008 4:43 AM
Multicast doesn't work - BT break it.
At the moment - for the vast majority of people outside the cities, there is no choice but to use an ISP that does not have actual physical equipment in the exchange.
Each user is connected over a virtual BT-owned pipe (nomatter who the ISP is) to the ISPs operations centre.
Each of these pipes are _completely_ seperate encapsulated internet connections.
The ISP pays an equivalent of some 60p/gigabyte of transfer (at peak times) for this service.
There is no way other than the hugely expensive route of putting equipment in each exchange for the ISP to actually make any bandwidth saving by using multicast.
What happens is that the multicast signals get to the ISPs operation center, then they are split into copies, and all the copies are sent to each user individually.
Even if you have 50 people in one street all watching it, there isn't a way for the ISP to use this knowledge - they simply can't inject multicast packets at the user end from some seperately bought multicast service, as this does not exist.
Posted by: Ian Stirling | March 9, 2008 3:06 PM
Another company, MediaMerx, has developed a platform to deliver high quality video over broadband to markets that suffer from international bandwidth bottlenecks.
ISPs can create customized video players, source content from around the web, add local content and license premium content.
ISPs and broadband providers can sign up for a trial here: http://www.mediamerx.com/index.php?section=betaprogram
Posted by: Michael Hall | April 21, 2008 6:50 PM
Do other online services pay ISPs for bandwidth like this? ITV and Channel4 have IPTV. Does Google pay BT for the bandwidth costs of YouTube? This all seems crazy to me.
Posted by: Dave | May 15, 2008 8:47 AM
This seems an odd critism to pass to the BBC.
I understand the cause and effect and the natural conclusion that someone has to meet the cost... 'BBC caused it, they meet the costs'
But actually how many people signed upto 'unlimited' deals with their ISP simply because they did not want any risk of restriction from online activities.
I think 7 years ago (first got online...) I was aware that the web needed to speed up, it seemed inevitable that before long most media would start to transist to digital and that ultimately the web would be the only network required.
So are we to believe that I had more foresight then than the ISP's who have been offering unlimited download?
It was only a matter of time before someone broadcast mainstream media over the web.
Posted by: David | September 11, 2008 3:40 PM