Measuring and Improving Broadband

In addition to my last post on universal broadband, I’ve noticed that another aspect of broadband infrastructure the soon-to-exceed-one-trillion-dollar-economic-stimulus-plan will tackle is increasing broadband speeds. The narrative goes something like this: most U.S. consumers are forced to make do with third-world broadband speeds topping out at a couple of megabits per second. Other countries deliver broadband speeds in excess of 100mbps, thus creating a competitive disadvantage that only government spending can overcome. (Does this mean that only government spending can help the Oklahoma City Thunder compete with the Los Angeles Lakers?).

But how, I wonder, are the speeds in the United States and other countries being measured? After I spoke with a number of analysts who cover broadband access, it became apparent that most, if not all, of these comparisons are between advertised speeds. Believe it or not, it appears that many within government actually believe that a 50mbps broadband service, for example, actually allows subscribers to download traffic at 50mbps.

I thought almost everyone knew that advertised broadband speeds are really an expression of the peak instantaneous downstream bit rates. Due to oversubscription in the design of broadband aggregation networks, average downstream bit rates are often as little as one-tenth the peak downstream rate. This isn’t false advertising and is, in fact, an economical way to build broadband infrastructure that delivers services consumers can afford. And the vast majority of data applications (Web surfing, e-mail, and so on) operate quite nicely on oversubscribed networks because they only consume bandwidth intermittently.

But this design philosophy begins to break down when streaming applications, video in particular, occupy a disproportionate percentage of bandwidth. That is precisely what is happening today. Simply increasing speed (again, peak downstream bit rate) will not necessarily improve broadband performance. If it did, we could improve performance without spending any money — simply increase the degree of oversubscription and, presto, broadband speeds can be doubled. Kind of like solving freeway congestion by increasing posted speed limits.

While peak downstream bit rates are not unimportant, improving the performance of broadband infrastructure requires investment in two areas:

Lessen the rate of oversubscription. This is actually very expensive and could result in broadband access charges significantly greater than those currently paid by consumers. Reducing the rate of oversubscription from 10-to-1 to 5-to-1 requires a doubling of bandwidth in the aggregation and backhaul network. Reducing it to 1-to-1 would require a tenfold increase in bandwidth. Nevertheless, investment in this part of the network yields far greater performance improvements for the consumer than increasing access speeds. For this reason, government programs designed to improve the broadband infrastructure should be focused on aggregation and backhaul networks.

Improve handling of video traffic. The design philosophy of the Internet (especially its protocols) clearly did not contemplate streaming video (especially long-form, increasingly high-def video). While data applications operate very efficiently and fairly, real-time streaming video is exceedingly challenging in a best-effort environment. (Those who believe trickle-down or progressive streaming methods get around this problem can read this). Many broadband providers recognize that the consumer goal is an Internet-video viewing experience indistinguishable from a locally attached DVD (even Blu-Ray) player. These providers thus have begun to add intelligence to their networks that delivers video traffic over dedicated “channels” that are separate from high-speed Internet (HSI) access yet share the same broadband access facility. In some ways this follows the strategy employed by content delivery networks (CDNs): sending high-value video traffic over dedicated network facilities separate from the best-effort (which means equally bad for all) Internet. This is another area in which government assistance would yield results far more apparent to the consumer than simply increasing downstream speeds.

So, yes, as we look at infrastructure investment in things that are a little more 21st century than bridges and highways, broadband is a good place to start. But focusing on peak downstream bit rates is simplistic and will not produce the desired results. Focusing on aggregation network bandwidth and service differentiation will.

Kevin Walsh is vice president of marketing at Zeugma Systems, which sells gear to allow broadband service providers to identify, monitor, manage and customize traffic flows on a per-service, per-subscriber level.