Comment: NBN multicast falls short

NBN Co's multicast IP product will become available for the first time next year, allowing service providers to send data out once, reticulate again. Telecommunications consultant Paul Brooks asks if the product has gone down the wrong path.

Last week I came away from NBN Co's briefing on the multicast product thinking "nice start – but it could do better". Questions from the floor indicated many others felt the same way.

The multicast capability, as explained by NBN Co, has been squarely aimed at facilitating internet protocol television (IPTV) based on the traditional subscription or "pay TV" business model. The shame is that the artificial and unnecessary constraints built in to the product prevent it from being so much more.

Multicast IP has been used across the internet since its dawning days, primarily for distributing live video and audio feeds of conferences and meetings and 'virtual whiteboard' presentations.

A Rolling Stones concert was broadcast over the internet through the 'Mbone' (Multicast Backbone) back in 1994.

The version planned for the National Broadband Network, unveiled last week and set for launch next year, requires every service provider to effectively purchase three components.

The first, the multicast access virtual circuit, must be bought at a minimum 20 Mbps capacity per premises, for $5 per month. This is mandatory for a multicast service, regardless of requirements.

For each premises, service providers can provision individual "media streams" of 3 Mbps and upwards, each of which is expected to hold more than one audio or visual channel depending on requirements.

NBN Co has already agreed to subsidise the service provider's the first 200 media streams per point of interconnect, with each additional stream costing $50 per month, per point.

These a provisioned within a third cost component - a wider multicast domain that is effectively the equivalent of the connectivity virtual circuit. It provides a collar of between 100 Mbps and a gigabit per point of interconnect, at $250 per month.

Multiply even just that last cost by 121 to get a substantial cost which will need to be recovered among subscribers.

Restrictive bandwidth

 The multicast service is provided on top of a conventional data service – you cannot have an IPTV service unless you also have internet (hopefully they don't both have to be from the same supplier!).

Apparently this is to allow for two-way communication over a multicast connection, such as changing channels and other data (from your remote control or set-top-box). However, there is nothing in the Layer 2 or Layer 3 multicast protocol specifications that requires a multicast stream to only operate in one direction.

Indeed, multicast IP is designed for the subscription protocol (IGMPv3 for those in the know) to go upstream towards the data source. The NBN technical design relies on this to get the subscription protocol packets upstream to the service provider. There doesn't appear to be a technical basis for this restriction, so why have it?

I hope there is a better reason than just being a product packaging exercise to capture those that are happy getting voice and data from a cellular network, and just want the NBN for television.

The problem

 These cost components are entirely built around the idea that multicast will be used for video and some audio. If your desired service does not need 20 Mbps, or you can't get $5 - $15 per month of retail value out of the facility, you are out of luck.

As a provider, until you grow to have several hundred subscribers in a serving area, you will probably be wearing a loss.

To make matters worse, the service provider is required to dimension their multicast domain by provisioning for the peak bandwidth of each stream, rather than the average bandwidth.

Video streams are highly variable in bandwidth over short time periods – a sports broadcast might average 7 Mbps but peak occasionally at 15 Mbps, while cartoons are highly compressible. Broadcasters use the benefits of statistical multiplexing every day to ensure the peaks of one channel coincide with the troughs of another and the total bandwidth of a collection of channels is much less than the sum of the peaks.

NBN Co's model will essentially guarantee that for every commercial broadcaster, roughly half of their multicast domain they are charged for will be forever empty. The more channels they offer, the more will be wasted.

Do you only want to receive standard definition television, which only needs 3–6 Mbps, or digital radio which needs less than 1 Mbps? To receive the occasional corporate address? To view or listen to a university lecture or the local council meeting?

As a provider, you might want to use multicast to efficiently distribute software updates and data information services, or send out something equivalent to the 'switch on/off now" messages used for off-peak hot water heaters that currently arrive through the mains electricity cable.

NBN Co has publicly stated it will cater for alternative business models in the future. But without proper timeframes and the current framework planted firmly in the idea of IPTV, more providers can forget about these uses.

The commercial internet backbones could not work out how or who to bill for multicast traffic, leading many to block multicast IP packets and effectively killing 'Mbone' during the 1990s. Primarily this was due to commercial and pricing reasons rather than technical – and it seems the future of multicast services over the NBN is doomed to suffer from the same issue.

Multicast is billed as a "platform for innovation" and it should be. But television is hardly innovative and television viewing is declining.

Unfortunately the current pricing model and unnecessary constraints will ensure it will not be used for anything more.