It is a truth universally acknowledged, that re:Invent is an overwhelming show in its every aspect. From the garish assault to the senses that is the Las Vegas strip to the tens of thousands of people sprawled across a canvas of hotels and conference centers to the hundreds of product and service announcements, the annual AWS event is impossible to summarize in any meaningful way. Realistically the only approach to it is to think of it as something akin to stream processing; rather than trying to process the flood of events individually as transactions, to let them flow by, considering and acting on them in the aggregate instead.
In lieu of an impossible to deliver summary, then, here are a few high level takeaways from this year’s re:Invent. As a caveat, this is less strictly focused on what AWS announced at the show than what it all means in the wider context of the market. For those looking for a more detailed discussion of specific announcements, look for a follow up RedMonk Slack chat.
When it created the cloud market in 2006, AWS ushered in a multi-year era of dominance for DIY-style infrastructure. Given a choice between primitives such as compute and storage or the higher level abstraction of early Platform-as-a-Service (PaaS) offerings such as Force.com or Google App Engine, the market opted for the former to the extent that AWS has gone from a few products considered toys by market incumbents to a $36B run rate and seventy-five thousand people descending on Las Vegas the week after Thanksgiving.
The meteoric financial success of Microsoft’s core realization that software was potentially an end, in the revenue sense, rather than the means it had been treated to up till that point changed the technology industry. For decades after the introduction of one of Microsoft’s two licenses to print money, Windows, companies attempted to emulate its success by protecting their software and charging for it.
AWS has had a similarly outsized impact on the market. The runaway success of AWS has meant that the industry it has come to dominate has grown in its image, which is to say based on and built from primitives. Whether cooperating with AWS, competing with AWS, or more likely both, vendors today are in the same business that AWS is: to sell the pieces from which enterprises will construct the architectures that run their respective businesses.
With rare exceptions – Amplify as pointed out, Elastic Beanstalk, or similar – AWS has studiously avoided delivering integrated solutions to market, and for good reason. First, the market has shown an insatiable appetite for the primitives offered. But less obviously, AWS is organizationally optimized for building autonomous, independent products and services – not broad, sweeping integrated solutions.
But a question looms: how sustainable is the overwhelming dominance for primitives-over-solutions? In a world of limited options and constrained architectural approaches, the answer would be indefinitely. That’s not the world we live in, however. The technology landscape today is highly complex, heavily fragmented and becoming more so by the day. One logical response to this growing complexity, both for individuals and the enterprise, is higher levels of abstraction. Higher level, more abstract solutions – once referred to as “Integrated Innovation” – require fewer decisions between competing services and projects, fewer integration points (read: points of failure) and fewer opportunities for vendors to point fingers at each other when/if things go wrong. Abstractions are not without their own costs, of course, but are a reasonable response to complexity.
Heading in to this year’s re:Invent, therefore, one thing to watch was whether we would begin to see signs of higher level abstractions. In general, the answer was no, but in offerings such as Braket, EKS on Fargate or SageMaker Studio you did see signs that AWS is aware of an appetite for simpler, integrated and fully managed solutions. Whether this trend continues for AWS and at what rate will be something to watch.
One of our beliefs at RedMonk is that predicting the future is often a trivial exercise; it’s the timing of a prediction that’s difficult. This came to mind with the announcement of AWS’ Graviton2 instances.
Nearly a decade ago, the prediction was made in this space that ARM would become a major server player. It seemed clear at the time that while ARM was generally thought of as a chipset for mobile or other specialized workloads, its power advantages among other characteristics would give it an opportunity as a more general purpose compute platform for enterprise workloads. Outside of some limited and largely failed experiments, however, ARM never took off as expected.
Fast forward to this year’s re:Invent, where Andy Jassy was discussing the ARM-based Graviton2 chips which were the result of AWS’ 2015 acquisition of Annapurna Labs. According to AWS, these chips are 7X faster than the first generation of Graviton chips, which is interesting. Much more interesting is the fact that the ARM-powered M6 instances are faster than the x86-based M5 instances – and in some cases, by a lot. Broadly speaking, AWS expects their ARM-based offerings to be 40% better on a price/performance ratio than x86-based instances.
This is important for several reasons. First, it’s an opportunity – as if it needed another one – for AWS to differentiate itself in market. Second, it’s more competition for enterprise chipsets, which pushes the market as a whole forward. Last, and perhaps most importantly, it should be a galvanizing force for ARM’s software compatibility story, perhaps the biggest inhibitor to more widespread ARM enterprise adoption. It was one thing when small startups like Calxeda promised price/performance benefits; it’s a different thing entirely when the industry’s dominant cloud platform provides fast ARM-based instances and uses precious airtime at its annual event to promote them as faster than the traditional Intel x86-based offerings.
All these years later, that original prediction might finally come true.
In life and in business, if there’s one consistent rule about leadership it’s that those who feel compelled to announce that they are leaders are not. True leadership is by general consensus, not unilateral decree. This fact does not stop would be leaders from loudly and repeatedly boasting of their leadership, of course, but it does make the exercise of sifting winners from losers simpler because real leaders have no need to boast.
In business, however, dominant leaders are marked not by their claims of power, or their lack of same. Truly dominant market leaders are better identified, rather, by their self-deprecation.
It was notable, therefore, to hear Andy Jassy make the case for Amazon’s lack of dominance. From highlighting largely forgotten failures such as the Amazon Fire phone to noting the continued existence and success of business such as New Relic or MongoDB in the face of products like CloudWatch or DynamoDB, he put particular emphasis on the notion that there is room for a variety of competition even in markets in which Amazon directly competes. In his words, “Simply because we launch a business doesn’t mean it’s lights out for other competitors.”
The ostensible context was to emphasize the importance of focusing on the customer experience: in his view, this is the differentiator, and how competitors resisting or outcompeting Amazon have been successful.
As anyone who has followed the history of this industry understands, however, there is another reason that market dominant players choose to downplay their market dominance, and that may have been at play here. In any event, such a rhetorical approach is consistent with a player comfortable in its position at the head of a given market.
Amazon’s continued emphasis on open source didn’t get quite the same airtime on stage that it did last year with the release of Firecracker, but behind the scenes the work continues apace. From the open source track at the event to work to improve developer outreach, communications and smoothing the process of upstream contributions, the Amazon Open Source team remains committed to improving its process and contributions across the board.
The challenge in front of the group is not, in some respects, dissimilar to the reputational rehabilitation that Microsoft had to do under Satya Nadella. Like Microsoft, Amazon had a prickly and confrontational reputation with open source for years – though not to the same degree as the software giant. Given that its representatives used to have to be assigned security details to attend events like OSCON, it’s safe to say that Microsoft has done a remarkable job turning around its own internal attitude towards open source software and, in so doing, developer attitudes about Microsoft.
Whether Amazon will follow in these footsteps remains to be seen, but the level of commitment from a personnel standpoint, both on staff currently and in the pipeline, certainly suggests that the company is committed to the exercise.
One of the time-honored traditions of vendor sports is market participants taking potshots at one another. This is never more true than in the analyst-only private sessions, in which company leadership is seemingly judged in part on their ability to bring the snark.
But even with that as background, the focus from Amazon on Microsoft as a competitor was notable at this year’s re:Invent. Much attention was devoted to elevating competitive projects like Linux at the expense of competition like Windows by way of vehicles such as Microsoft’s poorly thought out and PR-unfriendly licensing changes.
It wasn’t the criticism itself which was a surprise; in most cases it was legitimate and in every case was typical for vendor events. In the aggregate, however, it was notable to see a dominant player such as Amazon focus on a particular competitor to the degree that it did. Which brings us to the last major takeaway from the show.
Perhaps the most strategic question concerning Amazon is one that was rarely if ever discussed at the show, which is one of scope. On the surface, it absurd to question AWS’ scope given the volume of services it currently has in market and the velocity at which new ones arrive. And indeed relative to any competitor anywhere – Microsoft included, AWS offers a deeper, wider set of production services.
Depending, that is, on how one defines production services.
AWS has, for its part, most often defined this as post-deployment. AWS offers developer tools, of course: aside from the just announced Braket or SageMaker Studio, it has had tools like CodeCommit or CodePipeline in market for four years. Relative to its core IaaS tools, however, these receive comparatively little attention and fanfare. The implication is that such tools were responses to customer demand as opposed to core products, which is understandable. Given how dominant AWS is as a platform, where code originates is not a strategic concern.
The question is whether that might change, and how it might change. Which brings us back to Microsoft, if indirectly.
In June of 2018, Microsoft announced its intention to acquire GitHub. Four months later, the version control platform announced a new platform called Actions. Actions was and is versatile in its capabilities, but its significance was, as characterized by my colleague, “profound.” Setting its specific capabilities aside, its ultimate import can be summed up simply: GitHub was no longer an island.
Up until that point, GitHub’s role in the universe essentially began and ended with version control and closely related functions such as issue tracking and management. Anything beyond core creation and versioning was an exercise left to third parties, partner and otherwise. With Actions, for the first time GitHub had a direct path from version controlled code to a production deployment. It was a new product, of course, so it was a constrained and feature-limited path relative to the complicated and multi-faceted build, test and deployment pipelines GitHub users had constructed over the years, but it was a path nonetheless.
GitHub, and by extension Microsoft, were very careful not to play favorites or put their thumb on the scale; GitHub code could be deployed to AWS, GCP or elsewhere exactly as easily as it could be to Azure. But between the acquisition and the release of Actions, Microsoft found itself in the possession of:
- The largest corpus of source code on the planet
- The number two cloud platform and
- A bridge between the two
The companies have moved cautiously since, and have not sought to exploit the linkages here to the degree that the Microsoft of old might have. But for AWS, it might be worth considering whether that combination might present a strategic threat at some future point in time.
To date, Amazon has been able to dominate the market simply by virtue of being the biggest, baddest deployment target in the world – it hasn’t had to devote much attention to on ramps. If Actions evolves into an increasingly sophisticated on ramp to platforms, however, it is in a position to exert influence on travel to those platforms.
At present, the cloud platform is a decision of strategic importance and made accordingly; a scenario which in most cases favors AWS. But Microsoft’s Windows was once in a similar position. It was the biggest, baddest deployment target in the world, so Microsoft could feel free to pay less strategic attention to on ramps like Java middleware that might lead to other platforms like Linux.
The end result of those efforts was a decline in the strategic importance and role of the operating system; a shift which threatened Microsoft and changed the market landscape dramatically. If a similar commoditization is in the future for cloud platforms, it’s years away. But having clearly learned important lessons from Microsoft’s history already, it would not be shocking to see Amazon start to pay more attention to the world pre-production.
Their re:Invent punching bag Microsoft clearly is.
Disclosure: Amazon, GitHub and Microsoft are RedMonk customers. Google is not currently a RedMonk customer.