For several hours on Tuesday, big parts of the internet broke. Why? Because Amazon Web Services experienced an outage. Numerous websites and other web-dependent products were affected. Based on their bitter jokes, the businesspeople who rely on AWS to conduct and distribute their work were not pleased. "If you don't want your life to depend on AWS, the solution is simple. Just entomb yourself with an 80-year supply of canned food and water," said Techmeme boss Gabe Rivera.

As when the Dyn DNS outage took down an even larger portion of the internet, this incident prompted concerns about centralization. AWS may not be a single point of failure -- the majority of the world's websites were unaffected, including Twitter, where all the AWS grousing happened. But AWS does host a massive number of websites, especially startups too small to own their own servers. It's not a single point of failure but it's a significant one.

"The 'winner takes all' dynamic of the tech industry concentrates more and more power into fewer and fewer companies. That consolidation has implications for competition but also affects the resilience of the internet itself," Wired cautioned.

"[W]hat happens when any service gets so big that its tentacles touch the entire industry? Its failures become amplified to a destructive degree. In the case of AWS, that .01 percent of the time when your data isn't available means that over a third of the internet ceases to function well," Gizmodo warned.

Some observers have even concluded that AWS should be regulated as a monopoly. Atlantic reporter Robinson Meyer tweeted, "Hmmm, it seems like half the web goes down when Amazon S3 goes down. Let's check in on the status page for US antitrust law." Game developer Andi McClure expressed a similar sentiment: "Does anyone else see 'the U.S. Securities and Exchange Commission' in big blinking red lights[?]" when reading a list of websites taken down by the outage.

The view that Amazon's share of the market is unfairly large is not just espoused on Twitter. For one thing, the president of the United States has criticized the company, although he didn't bother to detail his reasoning. It's a critique that applies not just to AWS but to the company's dominance in its core business of retail.

In the beginning of 2017, the Yale Law Journal published an essay asserting that Amazon's go-to-market strategies are anticompetitive. "Specifically," author Lina M. Khan says, "current [antitrust] doctrine underappreciates the risk of predatory pricing and how integration across distinct business lines may prove anticompetitive." Khan argues that "Amazon's willingness to sustain losses and invest aggressively at the expense of profits, coupled with its integration across sectors, has enabled it to establish a dominant structural role in the market."

Khan's analysis, however, misses that Amazon's marketplace advantages, in both ecommerce and cloud services, are driven by user choice rather than resource monopolies. Amazon doesn't own all of the servers in the world, or even in a given region. People don't use Amazon to buy home goods or host a website because they have to, but rather because Amazon offers a superior and competitively priced product. And contrary to Khan's accusations that Amazon's pricing is predatory, Google and Microsoft are just as well-placed as Amazon to slash what they charge. Both did exactly that in 2014, which prompted a response from AWS.

AngelList investor Parker Thompson responded to the fears of over-centralization by tweeting sarcastically, "Remember before Amazon when your hosting provider was insanely cheap, had amazing APIs, and never had issues? Those were the days." In a reply, he added that "some of the people complaining should be old enough to remember how shitty ops life was in 2004."

In an email to Inc., Thompson expanded on his thoughts, pointing out that Amazon's competitors don't consider AWS to be a monopoly. "Companies like Google, Microsoft, [and] Digital Ocean are competing for the cloud market. Companies like Rackspace offer better managed solutions than they did when AWS was launched. And companies like IBM, HP, and my alma matter Pivotal are building software layers that seek to commoditize these cloud offerings and let companies move easily between [them]."

Thompson added that failing to plan for outages can make their effect much more catastrophic. "I can recall an earlier AWS outage where many AWS customers were affected because they relied on co-locating their services within one AWS datacenter, but Netflix stayed up because their engineering team had designed their software to continue working in this situation."

None of this is to say that AWS isn't beating its competitors. As the situation stands, it certainly is. In a February report, Synergy Research Group noted, "Amazon Web Services (AWS) is maintaining its dominant share of the burgeoning public cloud services market at over 40 percent, while the three main chasing cloud providers -- Microsoft, Google and IBM -- are gaining ground but at the expense of smaller players in the market."

Neither can anyone argue that centralization doesn't come with tradeoffs. It is simply common sense that when everyone relies on a handful of service providers, the biggest of those service providers going down will have an impact. But centralization has pros as well as cons. Parker Thompson told Inc., "There are obviously 'bad things' that can result from [centralization], like government monitoring and censorship, catastrophic outages, etc. There are also 'good things' that result from that, like the kind of technical advances, power efficiencies, and pricing that you can't get running web servers in your bedroom."

That's the rub. Hosting is hard, and AWS is usually very good at it. The service level agreement for Amazon's Simple Storage Service offers credits to users when uptime is less than 99.99 percent. Even the most expertly run service can't stay up 100 percent of the time -- it's impossible.

Veteran entrepreneur and tech podcaster Marco Arment pointed out, "Nobody can gloat during the S3 outage. EVERY hosting platform goes down. It's just a lesson for the young people who thought AWS didn't." He followed up with a warning that echoes Thompson's comment about servers in the bedroom: "Worth asking yourself if you build on AWS: If this goes to shit and you want to move elsewhere, how easy is that going to be?"

"My guess is that if the topic is uptime," Parker Thompson said, "Amazon (centralization) has significantly improved the situation. We just notice outages because now 'everyone' goes down at the same time. I'm not sure that's the end of the world."

Published on: Mar 1, 2017