Many have been citing the FCC’s National Broadband Plan (NBP) as conclusive evidence that that high speed broadband would turn into a cable monopoly with the universal deployment of DOCSIS 3.0.  Google in their reply comments to the FCC on Net Neutrality proceedings stated:

“As the FCC described in its National Broadband Plan, the current broadband duopoly is declining to a monopoly market, as approximately 75% of U.S. consumers “will likely have only one service provider (cable companies with DOCSIS 3.0-enabled infrastructure) that can offer very high peak download speeds.””

I managed to ask Blair Levin (head of the FCC National Broadband Plan) at last Friday’s “Future of the Internet” conference at Stanford for some clarification on this and managed to get it on the video below.  Levin not only dispelled any notion that the FCC made any such predictions of an impeding broadband monopoly, he also pointed out that both political parties have been stretching the meaning of the FCC’s broadband report.

The FCC basically looked at the currently announced deployment plans of all the broadband providers and observed that other than all fiber networks such as Verizon FiOS which will only cover a small percentage of homes, only Cable companies have announced 50 to 100 Megabits per second (Mbps) broadband services (peak speed) on a large national scale using DOCSIS 3.0 technology.  This is a far stretch from the assertions that the FCC is concluding that DOCSIS 3.0 will have a broadband monopoly and there are many problems with this line of thought.

For one thing, it ignores the progress being made on the copper telephone wire side when Lucent has announced successful tests of speed tripling technologies for VDSL2..  While 300 Mbps over copper may not be practical on any significant scale in the real world using less than ideal quality copper, the advances in VDSL2 technology are real.  It’s quite possible that the Telcos will soon be able to reliably triple their common 25 or 32 Mbps VDSL2 sync rates to somewhere between 70 to 90 Mbps using more advanced VDSL2 modems with vectoring technology.  It’s also important to note that these VDSL2 speeds are dedicated and not shared between homes on the last mile.  A relatively small amount of bandwidth is shared with Internet Protocol Television (IPTV) services at 4-6 Mbps per HD IPTV stream, but even 4 concurrent HD televisions all watching a unique HD channel would only average 20 Mbps of bandwidth consumption.

The other limitation of cable DOCSIS 3.0 deployments is the heavily shared nature of the network.  While it’s theoretically possible that cable networks can run at 5000 Mbps per node, it would require a radical change in the current cable business model and require the elimination of analog and digital cable television services which is highly unlikely.  For practical purposes and in the near term, DOCSIS 3.0 networks will have node capacities between 160 and 320 Mbps, and each node will serve neighborhoods with approximately 150 homes (assuming 30% broadband uptake on 500 homes).  That means realistically, the percentage of customers with 50 Mbps or greater must remain low or the system would be overloaded.  Figure 1 below shows how the capacity of cable compares to VDSL2 (FTTN) and something like Verizon FiOS (FTTH).

Figure 1 – How broadband technologies compare under load

Figure 1 compares the potential bandwidth performance of broadband technologies on the access portion of the network.  It assumes 10 homes sharing an FTTH node and 150 homes sharing a DOCSIS 3.0 node.  Also note that the total capacity numbers are based on “signaling rates” and actual performance will be 10 to 20 percent lower due to overhead.  “Average user activity” indicates the average per subscriber bandwidth consumption and not the percentage of activity on the entire network.  So if we have 20% of cable broadband subscribers at 10% link utilization level and 80% of the subscribers idle, the average user activity of all subscribers is 2% link utilization.

Cable operators will always be able to offer higher peak speeds than copper networks which means cable will have a monopoly on bragging rights over copper based broadband services.  That does not translate to a monopoly over any significant percentage of the broadband market much less the entire broadband market.  Telco operators will continue to be able to offer substantially higher minimum guaranteed speeds even under high neighborhood usage levels.  This is one of the main reasons Telco operators didn’t need to deploy BitTorrent throttling technologies like their cable competitors.

Will applications need 50 Mbps or more?

At the end of the video, Blair Levin openly questioned whether applications will drive a need for 50+ Mbps broadband.  As I’ve pointed out here and here, applications will always lag existing broadband capacity by an order of magnitude simply because of server-side economics.  This is why most applications require less than 1 Mbps even though we have tens of millions of broadband subscribers in the U.S. with more than 10 Mbps of broadband capability.  This is why applications in Japan are even narrower than applications in the U.S. because server-side bandwidth is more expensive in Japan.

Note: Server bandwidth is a better “pure” bandwidth cost metric because it’s typically unshared dedicated bandwidth and it generally refers to pure IP transit service for delivering data over the core of the Internet.  Client-side broadband bandwidth is a more nebulous number because broadband providers oversubscribe at very different levels, and different access networks have vastly different cost structures.  Most bandwidth metrics and studies mistakenly focus only on the client side of the bandwidth equation which explains the confusion about the dearth of high-bandwidth applications.

Applications aren’t narrow because they’re waiting for broadband to come up to speed; the opposite is actually true.  The real reason high bandwidth applications are avoided is because of the non-zero cost of bandwidth.  No matter how cheap bandwidth becomes, the cost will act as an upper limit constraint for applications.  As bandwidth continues to get exponentially cheaper, applications and content will continue to have larger bandwidth budgets.  This dynamic ensures that applications will always lag broadband performance rather than drive it.