This article is more than 1 year old
AWS won serverless – now all your software are kinda belong to them
Devs are lapping up Lambda, and this has already redefined the future of computing
Analysis "Serverless" didn't get a single mention in Amazon's quarterly earnings call. Neither did its poster child, AWS Lambda. Apparently it's not the sort of thing investors want to hear about, even if the developers that keep dumping cash into Amazon's pockets do.
In the latest quarter, AWS represented 73 per cent of Amazon's operating income, and serverless, perhaps more than anything else, represents the future of AWS's ongoing love affair with developers.
It also represents Amazon's best chance to completely dominate cloud profits for decades to come. Serverless is rapidly approaching "game over" and AWS is set to clean up.
This little thing called serverless
Introduced at AWS re:Invent in November 2014, AWS Lambda specifically and serverless generally have found fertile ground among enterprises. According to SlashData developer surveys, AWS leads the pack with 44 per cent adoption (compared to Microsoft Azure at 25 per cent and Google Cloud at 24 per cent).
Every other cloud vendor is a rounding error. But even this 19-point gap might be too generous to AWS competitors.
According to Cloud Native Computing Foundation survey data, the AWS Lambda lead is much more pronounced than the SlashData findings suggest: it's employed by 70 per cent of respondents using serverless tech. This is compared to 13 per cent who use Google Cloud Functions; 12 per cent using Azure Functions; and 12 per cent using Apache OpenWhisk (presumably there's some overlap in usage here as the total is over 100 per cent).
Serverless: Should we be scared? Maybe. Is it a silly name? PossiblyREAD MORE
Lambda launched in 2014 and its competitor cloud offerings appeared in 2016. While a two-year head start doesn't sound like much, "AWS really got people to pay attention to Lambda and – unusually for enterprises – start using it quickly," says Gartner chief of research Daryl Plummer.
One example is Expedia, which embraced Lambda early and often. According to engineering vice president Subbu Allamaraju, Expedia did "over 2.3 billion Lambda calls per month" back in December 2016. That number jumped four and a half times year-over-year in 2017 (to 6.2 billion requests) and continues to rise in 2018.
Just as importantly, AWS also leads in developer satisfaction. SlashData says Lambda has fewer developers who have evaluated but stopped using the platform (8 per cent total) compared to Microsoft and Google, each with more than 10 per cent reporting they've rejected these serverless alternatives. While that difference seems small, think about how many more users AWS has.
Actually, that's more a measure of developer dissatisfaction. When SlashData tracked developer satisfaction, AWS came out well ahead.
SlashData measured satisfaction across 13 different attributes per vendor and calculated an overall Net Promoter Score (NPS) showing how likely developers were to recommend their platform to peers.
AWS led on satisfaction (34) and NPS (32) versus Azure (satisfaction 27, NPS 28) and Google (satisfaction 25, NPS 24). Everybody else scored significantly lower than the leading three. "This looks like a market that will continue to be dominated by the cloud giants," SlashData said.
Perhaps it's obvious why AWS's lead matters, but most don't recognise just how much it matters.
World domination, serverless style
Leading Edge Forum's Simon Wardley, never one to mince words, helps to parse what a 70 per cent (or 44 per cent) lead means: "Let me translate that for you. Amazon is currently positioned to own 70 per cent of the future of ALL software." Developers, for their part, happily focus on writing business logic while AWS (or Microsoft/Google) handle all the server infrastructure. As Matt Wood, AWS general manager of Deep Learning and AI, told me: "With S3, DynamoDB, and Lambda, you can build apps without thinking about the underlying infrastructure."
This makes the developers dramatically more productive even as it makes the cloud vendors rich.
It also ensures that a developer building on a particular cloud will be unlikely to leave it. An app hardwired to AWS Lambda simply isn't going to be able to casually drop those Lambda functions for an alternative on Azure or Google Cloud. CoreOS founder (and now Red Hat exec) Alex Polvi detailed the reasons: "It's code that's tied not just to hardware – which we've seen before – but to a data center, you can't even get the hardware yourself.
'Lambda and serverless is one of the worst forms of proprietary lock-in we've ever seen in the history of humanity'READ MORE
"And that hardware is now custom fabbed for the cloud providers with dark fiber that runs all around the world, just for them. So literally the application you write will never get the performance or responsiveness or the ability to be ported somewhere else without having the deployment footprint of Amazon."
And yet developers will continue to embrace serverless because it makes their lives so much easier. Partly it's the new reality of dealing with lock-in. As Allamaraju suggests, rather than wring their hands over software or service lock-in, enterprises should "embrace techniques like service orientation, asynchronous and decoupled communication patterns, micro-architectures, experimentation, failing fast, tolerance for mistakes, chaos engineering, constant feedback and continuous learning."
Meanwhile, AWS will clean up.
Because of data economies of scale, Wardley explains, serverless is a true "winner takes (nearly) all" market. "As Amazon's serverless ecosystem grows, the more metadata it can mine, the faster its rates of innovation, customer focus and efficiency. Once it gets to around 2 per cent of the market then it's game over for all those not playing at scale."
For those thinking, "OK, that sounds serious, I'd better step up my serverless game," the stakes are too high, and the threat too imminent. Wardley goes on: "We can't be more than one to two years from Amazon hitting that 2 per cent of the serverless (i.e. platform) market. Which means, if you're going to counter play then NOW (not six months later) is the time to be dropping $5 to $10bn on that counter play."
The reason the stakes are so high is that we're talking about the platforms upon which nearly all enterprise software will be built. Serverless is a virtuous cycle for the winners, because efficiencies and data earnings get bigger and bigger for the dominant few, and lock out everyone else. Can all software move to a serverless platform like Lambda? No. But much, perhaps most, can, and that "most" will be enough to redefine enterprise computing for decades to come. ®