Splunk CEO Godfrey Sullivan said it in the fewest words: “Cloud is the new data center, and AWS is the best data center.”
Sullivan was part of a chorus of corporate chieftains and tech leaders singing praises to the No. 1 cloud at AWS re:Invent, which drew an estimated 13,500 attendees to Las Vegas last week. Enterprises from Johnson & Johnson to Major League Baseball to the AWS poster child, Netflix, touted their production cloud deployments. Dozens of startups, many native to AWS, explained how they’re cleverly leveraging Amazon infrastructure and services.
Yet does Sullivan’s claim that AWS is "the best" stand up?
Well, it’s certainly the biggest. Last year, Gartner estimated Amazon had five times the cloud capacity of its nearest 14 competitors combined. Even more stunning is the momentum: AWS claims to have more than one million active customers, with a revenue growth rate of more than 40 percent year over year, according to AWS chief Andy Jassy (2013 saw $5 billion in revenue).
Mind share is another indicator. Check out this chart from a recent survey of 1,700 respondents conducted by IDG Enterprise, a division of InfoWorld’s parent company, which shows AWS 13 percent ahead of nearest-rival Microsoft Azure in thought leadership:
More important, as AWS execs proclaimed over and over at the keynote sessions, is the sheer breadth of services. Many have been around for a while, such as autoscaling, load-balancing, Elastic Beanstalk for developers, Redshift for big data processing, and hundreds of others, not to mention the countless third-party services built on AWS by partners. But the highlight of re:Invent was the rollout of new and noteworthy services built by AWS itself.
The wave of re:Invent announcements began with a new MySQL-compatible database engine, Amazon Aurora, followed by a trio of application lifecycle management tools, the AWS Key Management Service, an EC2 service to manage Docker containers, the AWS Lambda event-driven compute service, and the AWS Service Catalog.
This bounty of old and new services, plus the general cloud notion of no longer having to worry about physical infrastructure, led an attendee I randomly met on an elevator to marvel: “If you’re a developer, why would you develop anywhere else?”
Indeed, the lead that AWS has is almost worrisome. For a variety of reasons, I am more and more convinced that the bulk of business computing will move to the public cloud over the next 10 or 15 years. Personally, I don't want to see one cloud achieve total domination.
A multicloud future
That won't happen for a number of reasons. For one, AWS is an IaaS provider that has evolved to become sort of a multi-PaaS provider -- but it will never be a SaaS provider like Salesforce or Microsoft. That's not Amazon's business, and SaaS remains the biggest segment of the public cloud market. After all, in our cloud future, SaaS will inevitably replace COTS software.
But almost everything enterprise developers create themselves -- apps to engage customers and partners, even software that embodies the core intellectual property of a business -- will ultimately live on IaaS/PaaS clouds that enable enterprise developers to build, test, deploy, and scale better applications faster.
AWS's huge first-mover advantage may sustain its lead position, but it won't wipe out the competition, if only because enterprises can't tolerate business dependency on a single cloud giant. The talk I hear from enterprise IT management is about "spreading bets" across multiple clouds.
All of Amazon's nearest rivals have their strengths: Microsoft Azure not only caters to the .Net and Windows Server world, it's keeping up with the latest cloud technology, including hot open source plays like Docker, CoreOS, and Kubernetes. Google has the longest experience managing huge server and container infrastructure at scale and has integrated Compute Engine and App Engine into an IaaS/PaaS whole. IBM, with its latest cloud announcements, offers a Cloud Foundry PaaS on top of SoftLayer with a rich set of services for building analytics and machine learning applications.
Tools to manage deployments across multiple clouds emerged a while ago and are gaining more traction. At re:Invent, I met with CliQr, a multicloud management startup backed by Google Ventures that can dynamically determine which clouds should run which workloads. But a number of others, notably RightScale, enable you to manage and optimize resources and costs across clouds.
For me, the most exciting part of all of this is what happens when an enterprise goes 100 percent cloud. A session on Netflix's embrace of microservices architecture, presided over by senior engineering leader Sudhir Tonse, was a master class on how everything changes when you develop only cloud-native applications. Netflix's ability to iterate and string together new apps from a multitude of single-purpose cloud services has already become the stuff of legend.
Could Netflix have achieved this on another cloud? No, because none of the others were mature enough when the company began reinventing its architecture four years ago. The rivals are catching up in capability, but they'd better pull out the stops or risk falling further behind.