If you’re like me, you travel a fair amount for work, and that means you spend time sitting in airport concourses waiting for your flight, waiting for your flight to be rescheduled, and waiting for information on alternate flights when your flight is cancelled. Bottom line: There’s a lot of latency built into airline travel. My airport waiting time has led me to think about how these experiences relate to what I know about the infrastructure in my line of work: It’s urgent that the infrastructure is constantly evolving.
The airport system in the U.S. today is largely based on a hub-and-spoke model that seeks efficiency for airlines and for the overall travelling public by using centralized hubs that connect to a large number of “spoke” cities. In many ways, that is exactly how the internet is designed today: Large centralized data centers are the core of the cloud, and data and services are delivered to their destinations from these hubs. Yes, there are a lot of efficiencies to this centralized model (in terms of the cost, since many of these large data centers are in remote locations where power is cheap), but there are some big inefficiencies to this centralized model as well.
I will get to that in a moment, but first, let’s get back to the airport because there are clear lessons at the concourse for how the internet needs to change and the role that edge computing plays. If a passenger wants to fly from Portland, Oregon, to Portland, Maine, for example, there are no direct routes to get from one Portland to the other. No way to have “chowdah” and “lobstah” in charmingly rustic Portland, Maine, and then hop on a single commercial flight to the eccentrically weird Portland, Oregon, for a free-range, organic, vegan kale salad and micro-brewed kombucha. To get from one to the other, you’ll need to go through a hub airport — like Newark, New Jersey; Chicago; or Denver — and that doubles the length of the trip. That turns what would otherwise be a five-hour flight into an all-day ordeal.
Yes, this centralized model may deliver some key overall efficiencies because there may not be massive demand every day for a direct flight between two locations like the east and west Portlands, but there are major inefficiencies for travelers who don’t want to go to and from non-hub cities like these. If you happen to live near a hub airport, you will be spared this built-in latency. If you to need to travel to a hub city, you will also be spared this built-in latency. But it’s a very different scenario if you don’t live in a hub city or aren’t travelling to one.
The centralized data center model behind the cloud today creates the same dynamic. If you happen to live or work in a city with a mega data center or major peering exchange, you are on the fast track for mobile computing. But the experience is very different for markets that aren’t data hubs. The services, content, and data you need might live very far away and have to travel a very circuitous route to get to you — the equivalent of a multi-stop trip with layovers, plane changes, and flight delays that add unacceptable latency to your mobile computing experience. For certain markets that are hub cities, you generally should have a great mobile computing experience. Just like it’s great that people in Atlanta can get a direct flight to most of the world’s major cities … but what about everyone in Nashville, Tennesee, and De Moines, Iowa, and the Portlands, to name a few?
The hub-and-spoke model seems to be an economic imperative for air travel today in order to keep flights affordable, give travelers more options, and fill seats so that flights are profitable for airlines. The same model made sense at one point in time for the internet and cloud, but a centralized model is not capable of supporting the growth of wireless devices, mobile computing, the IoT, and other trends that require near-zero latency in order to perform as designed.
What we need are more “direct flights” and “short routes” for web-based data and services in order to support the kinds of things that consumers and commercial companies are doing with their smartphones, tablets, IoT sensors, smart products, and other wireless devices. To accomplish that, there needs to be edge infrastructure that reduces the distance between users and what they want to access. Here, some models make more sense than others.
I obviously have my critiques of the hub-and-spoke model of airport travel, and not just because I don’t like layovers. But, there is a key thing airports get right that is instructive for edge computing: colocation. So much of the discussion of edge computing today is based on the equivalent of having every airline build its own local airport in hundreds of locations across the country. This go-it-alone approach to the edge would be like Des Moines suddenly having an airport for United, a separate one for Frontier, one for Delta, and then a dozen others. And then repeat that in Spokane, Jacksonville, and hundreds of other cities … that makes just as little sense for the edge as it does for airports.
What makes far better sense is to create environments where IT services, content, wireless and fiber companies, and all of the other stakeholders bring their assets and services together in a shared environment. This is colocation at the edge. That is a fundamental efficiency of airports today, and it’s a model that has worked exceedingly well elsewhere in the IT industry — a key part of the foundation for how much the internet has grown over the past two decades. Colocation at the edge is the only model that makes sense, both technically and in terms of cost efficiency, for doing edge computing at scale. Our industry doesn’t need a model that would work solely for one city; we need an approach that allows us to move forward rapidly, and in an economically viable way, to hundreds of locations.
Scale is the most overlooked topic in edge computing today, but it needs a brighter spotlight because it is the key to separating models that are viable for only a handful of deployments to models that will truly work for the geographic reach that edge computing needs to achieve in the next few years.
Here’s to wishing there was something like edge in the commercial airline business because there was just an announcement about my connecting flight and, let’s just say, the latency I’m experiencing is giving me plenty of time to write another article.