There are two facts about Amazon Web Services that are difficult to argue with. First, that the company was singularly responsible for ushering in what might be most accurately described as the current era of primitives. Second, that this was an entirely unsurprising development given the organizational structure of the company. It’s difficult to argue with the first point because of the immense, almost bewildering nature of AWS’ current catalog of primitives. As for the second, no organization has come close to duplicating the company’s ability to design and launch service primitives at speed, which certainly suggests that the organizational structure is optimized to produce same.
In recent years, however, as has been documented both here and elsewhere, AWS has shown an increasing willingness to consider and, tactically, embrace the alternative to primitives in abstractions. Beginning a decade ago with Elastic Beanstalk (2011) and accelerating in recent years with Amplify (2017), Proton (2020) and App Runner (2021), the physical manifestation of AWS’ intentions – its products – suggested that an organization built around the primitives model was increasingly interested in courting audiences less interested in Do-It-Yourself (DIY), custom infrastructure from scratch than higher level abstractions that traded configuration optionality and choice for a smoother experience and better development velocity.
Which is why all eyes were on re:Invent this year to assess where AWS was headed directionally (more RedMonk discussion on that here, with a guest appearance from Corey Quinn). Would the company continue its strategy of complementing its immense portfolio of primitives with new abstractions? Or would it revert to business as usual?
In what was widely regarded as a less frenetic re:Invent than in years past, AWS did not – with all due respect to Amplify Studio – break much new ground with respect to abstractions. Instead the slate of news was more consistent with older re:Invent shows, featuring a litany of announcements around new primitives from an update to its Graviton product line to the availability of OpenZFS to Amazon FSx to a slate of serverless announcements around its core database services.
As if to drive this point home, CTO Werner Vogels candidly described the strategy as “Primitives, not frameworks.” While talking about the dizzying array of services the cloud provider offered now, he laughed and blamed it on customers for asking for them. The timing of this was curious because it directly preceded the announcement of what can only be described as the opposite of a primitive in the aforementioned Amplify Studio, and several AWS employees argued that this presentation was not in fact about a primitives-only approach.
If that’s case, however, it was expecting the market to appreciate nuance, and historically markets in any industry have not been particularly adept at that.
All of which means that this re:Invent was a fascinating event to watch, and will be that much more important to consider years from now.
It has been clear for some time that the demand for higher level abstractions is not an academic theory but a market reality. If AWS’ own product announcements from Amplify to App Runner were not enough evidence of this, all that’s required is listening to what forward looking, thoughtful developers are saying, as in the case of Simon Willison:
If AWS had the developer experience of Vercel they would get all of my money for the rest of my career
— Simon Willison (@simonw) December 1, 2021
or doing, as in the case of Joe Emison:
Today, we are live on @vercel, having switched from AWS S3+Cloudfront+Lambda@edge. Ultimately, Vercel offers a much better overall experience for both developers and users.
(Still using AppSync+Cognito+Dynamo+Lambda for back end)
— Joe Emison (@JoeEmison) December 9, 2021
The developer experience gap – the idea that the overwhelming volume of available projects, libraries and services is as much a challenge as a boon – is, likewise, not particularly controversial in most circles. Discussions with enterprises, vendors and developers these days all point inescapably at this conclusion.
It is possible, of course, that the lack of news about the more focused, domain specific abstractions that AWS has seemed to favor in recent years is an artifact rather than an implicit statement of direction and intent. The past two years have been chaotic and disrupted for everyone, and AWS for all of its deserved reputation for operational velocity is presumably not immune to that. The leadership transition, meanwhile, from the long tenured Andy Jassy to Adam Selipsky, doubtless also came with its own challenges. It could very well be, in other words, that the 2021 re:Invent is just a blip, not a deliberate tacking back to their original course.
But it is equally plausible, on the other hand, that this year’s re:Invent was evidence of a political divide within the company. One of the questions about AWS in recent years, as demand for higher level abstractions has ramped up, has been whether or not the company’s organizational structure was capable of meeting that demand. Development organizations designed to produce a high volume of autonomous, independent services, after all, do not typically have equal facility to launch higher level experiences that represent combinations of these primitives. Press reports, meanwhile, suggest that models that differ materially from AWS’ historical base primitives approach have met with some internal friction. And, hypothetically speaking, if an AWS employee was of the opinion that the embrace of higher level abstractions was a distraction at best and a mistake at worst, it’s difficult to imagine a stronger signal to send to customers, partners and AWS employees alike than a slide saying “Primitives, not frameworks” featured prominently in one of the prime keynote slots at your annual event.
The big question, then, that cannot be answered definitively yet, is what re:Invent 2021 means about whether AWS can or will reinvent itself moving forward. Inarguably the cloud market leader did not take steps towards widening its market via abstractions, but it remains to be seen whether that’s a momentary pause or whether this event was the first baby steps in walking back from that model.
What is apparent, however, is that AWS’ inability or unwillingness to capitalize on the evident market demand with the alacrity that has come to characterize its behavior in other areas is creating a sizable market opportunity for companies more comfortable with abstractions than AWS, as Vercel is demonstrating.
Either way, the year ahead will tell us much about which path AWS is going down, which in turn will tell us a great deal about where the industry around it will be headed.
Disclosure: AWS is a RedMonk customer. Vercel is not currently a RedMonk customer.