Clouds are like airlines
If you’ve read Nic Carr’s book The Big Switch, you’ll have heard the analogy that cloud computing is like eletricity: we’re moving from computers to computing, just as we moved from generators to generation.
Certainly, clouds (at least, big public ones like Amazon Web Services and Google Compute Engine) have a lot in common with an electrical utility. They’re elastic, because you can change your demand and pay for what you use. They power other things, whether those things are stoves or web applications. And they’re ubiquitous, blending into the everyday until we take them for granted. But the electrical analogy falls apart in dangerous ways, and if we rely on it too much, we’ll make serious mistakes.
Back in May of 2010, I proposed a different model, which I presented at O’Reilly’s Gov 2.0 conference. I hadn’t thought much about it until recently, but Ian Rae suggested I write it down somewhere—apparently, since that presentation, the analogy has gone on to inform policy-makers in at least two countries. So here it is:
I think clouds are like airlines.
Consider that every nation has its own airline, even if it runs it at a loss. It’s simply one of the trappings of nationhood. Airlines are part of the national infrastructure. In bigger countries–particularly free-market ones–a few competing airlines can survive. The U.S. has several competing brands, for example. They’re heavily regulated, but they compete.
In every country, there are other airline business models. Take Netjets, a timeshare approach to air travel. Or Air Ambulance services to transport patients on critical journeys. Or regional airlines that feed a larger national carrier. Nearly every country has at least one other airline: its air force. In some cases, countries decide to eschew their own military, but instead form a treaty with another nation, and adopt some of that nation’s laws and processes.
Within its business, the airline is free to conduct marketing, have its own policies, and handle support. But there are guidelines within which airlines must work within a country, as well as globally.
Here are just a few of the similarities I’ve discussed with people over the past few years.
- Cross-carrier interoperability: Like clouds, airlines need to interoperate, to ensure that luggage and passengers get where they’re going. A set of agreed-upon routing codes and baggage tags make interoperability of passengers (workloads) and luggage (data) possible.
- Security and admissibility: Some things can’t travel in the cabin, and some can’t travel at all, for security reasons, just as clouds don’t want malware or material that violates copyright. The airlines are expected to conform to security norms—allowing air marshalls on board, maintaining employee credentials, and so on—just as cloud providers are expected to deliver security and confidentiality enforcement.
- Competitive cooperation: There’s a certain amount of détente and co-opetition among rival carriers: Competitors will honor one anothers’ tickets to help out in cases of interruptions.
- Peering points are regulated: Interaction between airline passengers happens in a sterile, regulated environment optimized for fast transit. Airports are simply the peering points of airlines, with frequent travellers having the equivalent of private peering.
- Consulting and professional services: The model works at the consulting and professional services level, too. We have self-service travel, but we also have travel agents, stitching together trips across multiple carriers and adding related services.
- Broad alliances: Just as there are carrier alliances like Star Alliance, there will be cloud alliances. Certain providers aligned around a particular stack (like Openstack or Cloudstack) will deliver a more consistent portability among themselves, and several providers may band together to offer global coverage (i.e. a German cloud could offer in-country storage of personal information when it is economically unfeasible for a global provider to run data centers within Germany.)
- Tiered services: Passengers enjoy different levels of service for different fees. The regulators’ roles are limited to ensuring that price collusion doesn’t take place, and airlines can adjust the price of a seat based on how far in advance it’s booked and what the demand is like. Similarly, a cloud can offer tiered pricing and discounts on reserved instances or sustained load.
It’s easy to see why this analogy appeals to policy-makers:
- Consider, for example, the definition of regulatory oversight of cloud security: there are perimeter-level security policies (what can go on board) and policies within the cloud (locked cockpit doors.)
- Or consider what happens when it’s time to define damage limitations on clouds that amount to national interests: The Warsaw and Montreal Conventions outline the liability of carriers if things are lost or damaged, or if there is an accident.
- Or consider how regulators might fine clouds for bad behavior. Today, if an airplane leaves the gate late, it costs the airline. This provides a potential framework for how to enforce guidelines on things like the speed with which you can get your data from a cloud when you try to leave, and the penalties a provider might face for non-compliance.
I like this analogy because clouds, like airlines, tend towards consolidation. There are economies of scale to be had, and efficient clouds, like efficient airlines, are essential to national security and a healthy economy. While each airline must work with those of other countries—that’s what those global loyalty programs are for—every airline has an allegiance to its home country.
The clouds-as-electricity analogy has always troubled me because it leaves little room for brand or competition. Electrons aren’t differentiated; there’s no penalty for “leaking electrons” and there’s little reason for tiers of service. If you lose electrons, they’re a simple commodity. If you lose luggage, people, or an airplane, there’s a far more complex resolution.
Ultimately, Carr’s electricity analogy is predicated almost entirely on the economics of clouds, without room for specialization and differentiation. As we decide what cloud providers will look like in the future, balancing competition and innovation with stability and national security, I think the airline industry is a good model to use.
Now if I can just figure out how to get airmiles out of them.