Of the ton of stuff announced at AWS re:Invent last week, my favorite item is a plain gray box called AWS Import/Export Snowball which is, to my knowledge, the first hardware ever offered by the world’s largest cloud provider.
Amazon calls it a “data transport appliance,” but it’s really a secure box of hard drives that holds up to 50TB. Just dump your data (for your Redshift data warehouse or whatever) into Snowball and ship it to Amazon via UPS -- the quickest, cheapest way to upload that many bits to the cloud.
Snowball is a metaphor for AWS’s ambition to devour enterprise IT one large bite at a time. So is the new AWS Database Migration Service, announced in the same breath by AWS senior vice president Andy Jassy. All that data goes one way, from your house to his, and AWS has plenty of capacity to crunch on it -- a staggering 10 times that of its nearest 14 public cloud competitors, according to the latest Gartner estimate.
Enterprise management is increasingly receptive to the public cloud pitch, and Amazon understands that: Many of this year’s re:Invent sessions and announcements were aimed higher up the food chain than the usual AWS developer audience.
But there’s a limit to how many existing systems enterprises will move to the cloud. In fact, in most cases enterprise “migration” to the public cloud is really about building new applications on AWS or Azure or wherever and phasing out on-prem applications as their lifecycles end. Up until now, the majority of new enterprise cloud applications have been customer-facing Web and mobile apps.
Your future is IoT and analytics
At re:Invent, Amazon made clear it believes that the next big wave of applications will be connected to the IoT (Internet of things). Amazon CTO Werner Vogels pulled the wraps off the core offering: AWS IoT, a platform that includes a multi-protocol gateway, a device registry, a rules engine, device authentication, and data encryption. Couple this with such existing services as AWS Lambda and Amazon Machine Learning, and developers will have a rich palette to draw on when building IoT apps.
The message to enterprises is an aggressive one: If you want to build for the future, build here. It didn’t hurt that GE, the company that has invested more than any other in the industrial Internet of things, took the stage to proclaim that it planned to run no less than 60 percent of its workloads on AWS.
As both Jassy and Vogels noted, there's a lot more to IoT than a bunch of connected devices or sensors. The true value is in the analysis of IoT data for continuous product, process, and business optimization. AWS already has a number of backend analytics services for that, including AWS Elastic MapReduce running Spark (plus dozens of third-party offerings). Last week's re:Invent introduced both Kinesis Analytics for time-series analysis of streaming data, and Amazon QuickSight, a full-blown SaaS BI application.
Perhaps more than any other re:Invent arrival, QuickSight highlighted the seriousness of Amazon’s enterprise ambitions. For the first time it will put a glossy, Amazon-native presentation layer for executives on top of AWS’s vast compute capacity and catalogue of services. We won’t be able to review QuickSight until 2016 when it becomes generally available, but you can bet the announcement stirred unease among incumbent BI providers.
The sheer breadth of services now offered on AWS -- along with an auto-scaling infrastructure conducive to microservices architecture -- can make you wonder why developers would want to build anywhere else. Plus, as Amazon senses, increasing frustration with the cost, complexity, and liability of maintaining one's own IT infrastructure has many in upper enterprise management yearning to "get out of the IT business."
Amazon plays to that yearning, but it's a bit of a bait and switch. AWS is a platform. Yes, you shift the burden of running physical infrastructure and maintaining services across the ecosystem to Amazon. But along with developers who know AWS, enterprises still need admins to keep everything running smoothly in the cloud at scale.
At the show I spoke with Michael Liebow, a managing director for Accenture, who believes enterprises need cloud admins who focus primarily on cost considerations -- making sure developers select the most cost-effective array of services, shut down instances when they're done, and so on. A little later, a representative of the cloud management provider CliQr told me he's still seeing enterprises repatriate applications built on AWS to their own servers due to the high cost of running cloud applications at scale.
AWS has an ever more amazing array of services and an 81-percent growth rate that any tech vendor would die for -- but let's keep things in perspective. Andy Jassy says that AWS is a $7.3 billion business. Oracle takes in more than five times, IBM more than 11 times, and Microsoft more than 12 times that revenue. All three incumbents have aggressive cloud initiatives that include analytics and IoT tooling. Certainly they will be able to hang on to many enterprise customers and usher them into the cloud era.
Microsoft and IBM in particular have credible hybrid cloud plays, whereas Amazon does not. Sometimes running your applications in the public cloud is more cost effective, and sometimes it isn't. A hybrid cloud gives enterprises the opportunity to make that choice.
Amazon has an enormous head start and the options available on AWS stand head and shoulders above the rest, but you can't simply upload IT or send it off in a panel truck. Enterprises still have lots of complex decisions to make as they move to the cloud. Based on what they need, along with the politics of existing vendor relationships, they'll find the cloud that suits them.