The earlier blogs in this week’s series looked at the increased demand over time for broadband speed and usage and concluded that the continued growth of broadband demand will ultimately put great stress on last-mile networks – to the point where, eventually, fiber becomes the only viable broadband technology.
But what about middle-mile fiber routes? Middle-mile networks have gotten faster over time. In the 1990s, when the predominant last-mile technology was dial-up Internet, the predominant middle-mile technology was a DS3, which delivers 45 Mbps. As millions of people started using broadband, middle-mile networks were upgraded to 1 Gbps lasers and 10 Gbps lasers. For the last decade, the predominant technology for middle-mile has been 100 Gbps lasers. Recently, middle-mile fiber construction and upgrades are using 400 Gbps lasers. Vendors have already developed and field-tested terabit lasers with a speed of 1,000 Gbps.
One of the primary conclusions in an earlier blog is that 25 years from now, our broadband networks will likely be carrying at 12 to 15 times the volume of data that they carry today, and probably more.
Unfortunately, there is no easy path to upgrade middle-mile networks to keep up with the expected future demand. Consider the following chart showing the capacity of middle-mile lasers, starting with the 100 Gbps laser that is used on most middle-mile routes today.
Unfortunately, faster lasers can’t alone satisfy the future demand for increased middle-mile bandwidth. Achieving faster speeds on middle-mile routes is going to require a lot of replacement of existing middle-mile fiber – to fiber that is clearer and that has fewer microscopic impediments. A lot of the fiber in use today can’t handle terabit or faster lasers.
Just like with coaxial cable networks, many of our middle-mile routes are also aging. We’re already seeing that some of the middle-mile routes built in the 1980s and 1990s are aging and deteriorating. The fiber built in those decades was not as clear as today’s fiber, plus we used construction techniques that stressed the fiber, which eventually results in microscopic cracks that impedes the light signal. We’ve developed both clearer fiber and better construction techniques, but those improvements are too late for the older existing fiber routes.
We are probably going to have to use a combination of three strategies to handle the middle-mile demand over the next 25 years:
- Rip and replace current fiber to higher-quality fiber to be able to handle terabit or faster lasers.
- Build a lot of new fiber alongside existing fiber routes to handle the increased capacity.
- Employ strategies for reducing the demand on middle-mile networks.
Realistically, it will take a combination of all of these strategies to handle the future expected demand. Some of the demand will be handled by replacing existing fiber routes. Some will be handled by building new fiber alongside existing middle-mile fiber routes. But a lot of the solution must come from reducing the strain on middle-mile networks.
There are several strategies that can reduce traffic on middle-mile fiber routes.
- Edge Computing. Starting around 2010, there was a national trend to move data processing to large data centers. However, it’s become clear that some computing functions are better handled at the edge, meaning close to customers. As an example, it’s far more efficient to build a small data center at a smart factory so that the data processing can be done locally to give real-time instructions to manufacturing machinery. It’s also smarter to put the data processing in the factory so that the factory doesn’t shut down if there is a broadband outage in the network connected to the factory. If a factory changes to an edge computing network, there is a significant decrease in the amount of bandwidth being carried over a middle-mile network. ISPs don’t have much power to convince users to migrate data processing to the edge other than perhaps to convince large broadband users that this is a better configuration.
- Caching. This means storing data close to users. The best example of this today is Netflix. The company has over 80 million customers in the U.S. and Canada today, and a large percentage of those folks watch the same set of popular shows. Netflix has been caching content closer to customers by placing servers with large ISPs in larger markets, where they store copies of the most popular shows on the network. This means that a lot of Netflix content is generated locally on an ISP’s last-mile network and doesn’t need to use middle-mile. Netflix sends one copy to a ISP rather than sending large numbers of individual video streams.
- Peering. This means carriers swapping broadband traffic to each other rather than sending all traffic to the major Internet hubs. Most large ISPs already have direct connections to the biggest bandwidth users – companies like Google, Netflix, Microsoft, Amazon, and Facebook. But peering can go a lot further. Peering today is generally only done by the largest ISPs in the largest markets. The way to increase the use of peering is to establish many new peering sites around the country. Far too much traffic is exchanged today at the giant carrier hotels located in major cities.
The bottom line conclusion is that our current middle-mile routes are inadequate to meet future demand that will increase broadband traffic by a factor of 15 times or more. While the increase in last-mile broadband demand can be handled by faster fiber lasers, the existing fiber on most middle-mile routes cannot handle terabit and faster lasers. This means we’ll have to replace a lot of middle-mile fiber while also upgrading electronics. There is a lot of questions about who might fund such a massive upgrade. And even if the major middle-mile routes are replaced, most middle-mile fiber is regional. My prediction is that within 15 to 20 years we’ll be having a lot of discussion about the impending collapse of middle-mile networks and carriers will be begging the FCC or Congress to provide a huge subsidy program to upgrade networks.