It’s that time of year — AWS re:Invent. The time when tens of thousands of developers, product managers, operations engineers, venture capitalists and industry experts congregate in Las Vegas to discuss the future of the cloud computing behemoth that is Amazon.
I was fortunate enough to present center stage at the Venetian on Tuesday morning with Paul Underwood, an AWS Solutions Architect who works closely with young and innovative startups. We shared a vision of the serverless future, how the transition ought to happen, and how what we’re building at Polybit — stdlib: A Function as a Service Software Library — helps developers and companies build, maintain, organize, share and scale services in this new frontier.
AWS: The System Architecture of the Web
In the same way that JavaScript developers don’t worry about the memory addresses of the functions they write, future cloud developers in a serverless world will never have to worry about the physical locations, geography, VMs or containers in which their code executes. They will build services that just work. Period. Scale will cease to be a concern. Some of these services will be shared within an organization, others with the world as a whole.
In a sense, you can consider providers like AWS as system architectures of the web. Software-configurable infrastructure provides us with a wide array of options to accomplish a specific task. The breadth of these infrastructure offerings presents a large opportunity for opinionated software to help create an abstraction layer on top that drastically simplifies large, at-scale software development.
Parallels to Early System Development
Viewing cloud infrastructure from this perspective, we can draw parallels to the evolution of IBM and Microsoft for a guideline as to how creating abstraction layers (such as an operating system) on top of a compute layer impacts the progression of technology within the industry. I would argue that the future of cloud computation is not rooted in offering more and more specialized infrastructure services and not about the configuration of cloud-based systems in any combination you desire. The future is reductive, in a sense— it’s about enabling developers to forget about the underlying infrastructure implementation details altogether. No hardware, no metaphor for hardware like containers, just code.
Over the next decade, cloud compute volume (and competition between AWS, Azure, and other players) will continue to increase dramatically, but spending on cloud services will shift up the value chain to the application layer, in the same way we previously observed with Microsoft on top of IBM — focus moved towards user-friendly operating systems that enabled businesses to move faster instead of the computer layer itself. Margins and profits on infrastructure as a service will trend towards zero as computation becomes commoditized, with industry spend rewarding players focused on enabling usability over raw system configuration.
The Maturing Cloud Market
The shift in focus towards user experience on standalone systems in the past can be used as a map to help us predict how the cloud compute market will change. Infrastructure providers simply do not have the vision and maneuverability to execute on building the next layer of abstraction. It is a challenging task for larger players to shift up the value chain alone, because the design-by-committee mantra that dominates established corporations is not a suitable environment to foster the growth of the first iterations of these “operating system of the cloud” abstractions.
As the serverless space matures, we will see startups, driven by talented teams focused on building simple developer products, dominate the field. This has been, and will continue to be, the status quo of the emerging technology market. Think of the successes of Dropbox or Stripe, but applied to cloud computation as a whole instead of a specific business vertical.
The infrastructure providers that pave the way for these startups to flourish will ultimately set themselves the best for success in the future. Providers that spread themselves too thin trying to dominate both the commodity and usability layers are in for a rude awakening — a death by a thousand needles, built atop their competitors. AWS, though more than capable of outspending startups, has historically shown an awareness of this, with a certain willingness to “play ball” and encourage innovation on top of their platform — Heroku being one of the largest successes.
Looking Towards the Future
Companies like Firebase and Parse acted as pioneers in the cloud services space, focused on developer experience with infrastructure as a commodity. I believe we will look back on them as the early indicators for the upcoming “Cambrian explosion” of developer-focused ease-of-use tooling and abstractions in the cloud infrastructure market.
Our data certainly suggests this is the case. Within weeks of launching the latest version of stdlib we saw developer adoption (measured by compute time on platform) skyrocket. After months of iteration and nailing down the best serverless workflows, our growth chart, as measured by platform usage, looks like this (Y-Axis labels obfuscated);
stdlib Compute — September to December 2016
Thousands of developers have registered for our platform, with hundreds actively creating services in just the past few weeks. This is only the beginning. The usage of the serverless developer tools and frameworks today is just a drop in the bucket of industry-wide transition that’s underway.
How You Can Help Build What’s Next
Keep pushing forward. Infrastructure management is old hat. Feature delivery without operational management is where the industry is heading. Maintain a key focus on developer experience as you build and experiment with new technology.
The shift to serverless architectures is the first step in a much larger transition in how we think about interfacing with the cloud. Our vision for the future is relatively straightforward — the computational resources of our planet will be accessible, seamlessly, to any developer worldwide, without needing domain-specific infrastructure expertise. This will not only have a fundamentally disruptive impact on how people think about developing software, but also how people think about forming business entities in the new “API economy” altogether.
At Polybit, we’re building this future. We’re excited about what’s coming next, and you should be, too. We’re thankful for our wonderful developer community, the kindness of AWS to give us a stage in which to share our ideas and vision, and the reception we’ve received. The act of creating gigantic, globally-scaled platforms in the cloud is about to become an order of magnitude easier.
Keith Horwood is the founder and CEO of Polybit, where he’s building stdlib: A Standard Library for the Web. The team includes talented employees, investors, advisors, and mentors from AngelPad, YC, Google, Heroku, and AWS.
You can follow StdLib on Twitter, @StdLibHQ, or follow Keith directly, @keithwhor.