Those of us at io Networks thought it would be prudent to share some thoughts on the issues with NBN network & Netflix, illuminated by the ABC Network.
The ABC posted an article in July that complemented a piece on 7:30 that same night that touches on the capacity and throughput issues of Fixed Wireless that wrongly blames application providers for congesting networks, rather than the other way around.
Throughout the debate over the NBN, from its inception, planning, initial deployment and ultimately re-engineering, the political and technical merits of what was on offer never came even close to being to one another. It’s initial scope and cost were a primary target for conservative politicians, running high on the success of other campaigns to highlight “overspending” by progressive governments. Regardless of how important a complete overhaul of Australia’s aging, messy and monopolistic fixed line communications was, the discussion never came back to its benefit for the common good - just its cost.
When in actuality, an NBN, done properly, would have been a boon for either side of traditional political fault lines. It benefited both households and business equally. It would have boosted the economy in areas of the country where physical distance was a problem. It would remove many of the problems facing communications across a wide range of areas - mobile black spots, regional backhaul, the story state of most copper wiring, lack of reliable internet and phone services for business etc. But that wasn’t the story told, and the reactionary nature of both Australian politics and most of the non-tech media towed one line - it was expensive and it would be obsolete.
So what happened is what most of us in the industry expected would happen - the network would be designed to save money on the surface by cutting corners, but instead still hemorrhage money because there was not a consistent long term plan to account for growth over the time of the rollout. This is not a Netflix or streaming problem, as NBN Co recently floated to RSPs, as congestion caused by an explosion in streaming video could easily and should have easily been anticipated in the very early days of planning. Video has always been a big part of both local and wide networks since the mid-nineties, it was inevitable as storage and bandwidth speeds increased that it would become a huge part of a global network,
This lack of planning for capacity is a problem throughout the industry and this is thanks to a focus on short term costs over long term projections. Every single network will require more resources as it grows, and even a relatively static customer base will still require scaling over time. A network like the NBN should always be designed with growth in mind, taking into account that it was and is designed to facilitate the needs of 25 million Australians right now, and into the future. As both a public and legislated default Layer 2/3 network, it has a duty to be accurate in its projections and not be shocked by the demands of upper Layer applications.
Netflix is a temporary problem in the way that funding for the network is a temporary problem. Large scale networks that provide lock-in require a plan for long term survival. Netflix will come and go, along with a host of other streaming providers - right now it’s video for films and TV, in a few years it will be real-time games off the cloud, and in another, it will be a full-scale replacement of a local workstation. In 10 years time, we effectively could expect a 24/7 stream of content over IP networks in the same way a TV might run Free To Air or Cable News - across multiple devices in the same location.
This is coupled with NBN Co’s use of the CVC model, which they have recently attempted to moderate via wholesale bundling. This is a heavily regressive model for moderating congestion on a network, especially one that is as crucial as the NBN - very few national networks of a similar scale or type (such as NZ’s Chorus network) use CVC as it tends to force location rationing of bandwidth by RSPs, makes higher speed connections much less profitable, and increases user discomfort. Removing CVC as a method for revenue would drastically improve the performance of the network as a whole, putting amble pressure on both sides of the layers to stop blaming each other.
Right now, the core of the NBN network is well designed for scale - as the vast majority of it is reasonably centralized to the POIs, using standardized equipment that can be upgraded to suit. The major problem is that the last mile for roughly 40% of the network’s users is choked. Early plans, which may not be currently active, showed that nodes were utilizing 1Gbps SFPs rather than 10 or even 100, which would dramatically and pointlessly congest those nodes over the first 5 years. This is an easy and reasonably inexpensive fix, but the wider problem is sorting out a longer term last mile solution.
Fixed Wireless and Satelite were always challenging solutions for rural and regional users, but the issue was complicated by the lack of effective scalability. Both technologies are expensive and restrictive by design, they are primarily used as redundant links rather than primary, and scream of an “it would be fine for now” scenario. While it is possible that technology will improve, the upgrade and replication process is far more expensive down the track. In most regional cases, extending fibre or using traditional 4G technology would have been a much better, and easily scalable option. Very remote locations could then be carefully evaluated on a case-by-case basis.
On top of this, a lack of technology sharing with existing providers, both public and private, force NBN Co to overbuild infrastructure where it isn’t required. Most areas that sit on Fixed Wireless were done mainly for cost-effectiveness rather than a particular lack of suitability for fixed line. Most Australian towns tend to build housing around a city center, which tends to house the local telephone exchange. Converting this exchange to what would effectively be a large node would negate the need for Fixed Wireless and setup the town for the next 30-40 years. This has already been the design in a number of NBN enabled Tasmanian towns and others around the rest of the states.
Finally, throttling traffic is very rarely a good solution to limiting traffic use by users. Users will inevitably be less satisfied with their connections as a whole and will dispute the method in which this network is policed and metered. Net Neutrality works on a performance level as well as a political one (freedom of information as it were) because it requires less stress on packet examination. Charging users for different types of traffic, as opposed to caching or unmetering it, simply forces users to find effective workarounds, such as using VPNs or other tools that effectively slow down networks and create more discomfort overall.