Attracted last week to what we considered a contentious statement that IPTV is faced with an imminent bandwidth bottleneck, we called Continuous Computing and asked what on earth it was on about. Surely IPTV needs a backbone which is just 3.5 Mbps of bandwidth for each TV channel that it supports, call it a bouquet of 150 channels and it is a mere 500 Mbps for SD channels, with the access network needing how ever many multiples of 3.5 Mbps as consumers want concurrently running TV sets in their homes.
So what was all the fuss about? But talking to Mike Coward of the Continuous Chief technology Officer, he painted us a picture of a disjoint between Access Networks and the Core, with bandwidth shortfalls at the backhaul looming large on the horizon for every telco that invests in IPTV.
Continuous Computing began life in the VoIP business in 1998, but today finds itself in a multiplicity of internet business including Ad Insertion systems, Quality of Service Management and internet traffic shaping.
Coward argues that all you can eat broadband plans are going to have to come to an end, that there will be widespread introduction of traffic shaping, or traffic prioritization, and that with the imminent uptake of network PVR technology, the link between Access Networks and the Core would be assaulted with a whole new level of bandwidth requirements.
Continuous provides its traffic shaping devices, superfast packet processors that allow the inspection of every IP packet, to allow for prioritization of applications, as an OEM component to most of the major telecommunications equipment manufacturers, including Cisco and Alcatel and many of the smaller specialists. This type of packet processor can inspect, modify, tag, adjust and forward packets at extremely high speed.
Coward senses that regardless of the legal position in any given territory over net neutrality, Telcos will need to apply multiple strata of priorities all the traffic going through their currently bottlenecked backhaul for Access networks.
‘There just isn’t enough bandwidth to service everything in real time,’ says Coward, ‘ so Telcos are inspecting every IP packet near the edge of the network, and identifying the applications that are being run, and applying priority to them. Some of the lower priority packets just have to be thrown away,’ he argues.
By this he is not talking about data from Google deliberately being given less bandwidth than any other best effort internet web page, but different applications like P2P networks that are congesting networks with gigabyte file sizes are certainly one of the best applications to give low priority to, he said. ‘No one is going to sue you if you slow down an illegal file sharing application,’ said Coward.
We can argue about the morality of how you set a policy for traffic shaping, but he argues that telcos just can’t afford NOT to set such policies, and that in the future it is going to get a lot more congested.
‘The connection between access networks and the core is moving rapidly towards fiber, and bandwidth has always moved faster than Moores law. Networking bandwidth grows more rapidly than storage, and storage capacities, in percentage terms, grow more rapidly than processing power.
‘But there are a number of challenges ahead, such as nPVR, which will mean that instead of just 150 channels being sent over a private MPLS network, for an IPTV service, each Access pipe will have its own content, completely different from the next, which will be okay for the access network, but not for the backhaul that feeds it as this will need to be 100 Gbps even on the edge of a network.’
This is similar to what Alcatel calls Wirespeed, which means that as ‘always on’ applications such as IPTV emerge, the entire network needs to be able to operate to support the full capacity of the Access network, that part of the network attached directly to customers, when running at full speed.
Traffic shaping is of course one way to get around this in the short term, for any Telco that finds their network saturated. The difference between QoS management and Traffic Shaping is that in QoS systems, you are dealing with data that a service operator owns and that is initiated on its own network, like IPTV packets. It is already tagged, and parts of the network have been specifically built to be fast enough to sustain that data. In traffic shaping the telco has to look at all the unpredictable data traffic that people use and take decisions in real time about which stuff is the most important to deliver first.
Coward’s point about nPVR is probably right. There is a sense in the pay TV community that somehow nPVR is an inevitable technology waiting to happen for telcos and for cable alike. It will mean eating way at the $200 SD set top or DVR and the $450 HD set top or DVR, which takes up almost 50% of the cost of connecting each new customer. And yet this same technology is responsible for taking away a considerable number of viewers of advertising, something that is hurting the broadcast and cable TV networks in the US, denting their Upfront revenue commitments, and lowering total advertising revenues for the industry.
nPVR would cost just a few $ for each household, and provide much the same capabilities. It would also provide technology so that advertising could not be fast forwarded, which is a real bone of contention.
At the present the legal position in the US is far from clear over nPVR because content owners have just won a court case against Cablevision saying that it cannot use nPVRs, a decision which is being appealed to the US Supreme Court. But in the end if the cable operators are willing to pass some of the gains to the content owners, this position can simply be negotiated.
A PVR or DVR which can read your encrypted EPG is far easier to use than one that has no insight into what programming is coming, which is why TiVo does so well in the US, because it includes EPG services. Most consumers will take the equipment that is provided by the pay TV operator, and if Comcast, TWC, Cablevision, DirecTV and EchoStar all stopped offering a DVR which decrypted and used the EPG stream, DVRs would hardly sell at all.
Also there have been breakthroughs in the way that pay TV services offer VoD these days, with all the successful VoD campaigns offering a high degree of subscription or bundled VoD programming, and a lower proportion of paid VoD. This will mean that over a five year period VoD as a percentage of viewing may move from 6% or 7% now, if you include delay TV such as DVR, to over 70%. That just can’t happen if the backhaul for the access network does not have enough bandwidth to deliver a separate set of video services for each household.
When IPTV was originally introduced it was mostly by new telcos who got their bandwidth from unlit fiber in the ground, excess left over from the excesses of the Dotcom boom. Later, as tier 1 telcos joined in, that bandwidth came from existing bandwidth that was being used to support telephone exchanges. Okay there has been a lot of build out, as public capex statements from AT&T and the like will support, involving $ billions, but if IPTV operators want to extract the bandwidth they need to offer personalized video services to customers, including nPVR, VoD, personalized EPGs, personalized advertising, and they try to do it in the same way, there will be too little traffic left over for everybody else.
Of course there are arguments that none of this will happen. Pause TV, the idea where you put a $40 chunk of Flash memory into every TV that will allow you to pause TV and use it a little like a DVR, is in an advanced state and will emerge in 2008. And customers may churn dramatically if their DVRs are taken away from them, having already gotten used to avoiding advertising. Also addressable advertising, that which is most appropriate to the family, could be sent to a home and stored on a private portion of a DVR, sent once and played many times, rather than streaming it each time it is required.
And the rise and rise of over the top video strategies, which send Adobe Flash Video or other programming, from the internet over a broadband line, and then dropped onto the TV, would legislate against much of this change, but then again Telcos with advanced traffic shaping strategies could strangle this at birth, borrowing the bandwidth this needs for their own services.
This in turn is giving rise to new types of adverts, a little like the internet interstitial advertising, such as video pop overs, which come as part and parcel of the Adobe Media Player and which were last week launched on YouTube.
But Coward’s experience is that even if other options arise, the telcos themselves are ready to invest in this type of technology on a grand scale. ‘There are a new wave of platforms around IPTV, all based on the standardized Advanced Telecom Computing Architecture (ATCA) which make for cheap blades for very high bandwidth processing.’ There are now more than 100 companies offering products around ACTA.
Continuous buys specialist hardware off the shelf for its product line, based on the 64 bit MIPs processor, which Coward thinks of as commodity silicon. ‘People used to offer packet inspection class devices in the enterprise that were extremely expensive. These were all custom built and it took three years to get a new application into the market. Now that these commodity chips are standard and ready, new Telco services can come to market using packet inspection, in under 12 months. In the IPTV market if you take longer than 12 months to come to market with a new idea, it’s too late and the market has overtaken you.’