AWS CEO Unleashes Several New AI, Container And Database Services, Including Amazon's First Managed Kubernetes Platform
AWS introduced its first managed Kubernetes service Wednesday at the re:Invent conference in Las Vegas, as part of a flood of new solutions powering databases, democratizing machine learning, and delivering broad capabilities to developers looking to implement video and audio recognition.
Amazon Elastic Container Service for Kubernetes (EKS) is a managed container service that finally adopts the Kubernetes orchestration technology that the industry has standardized on over the last year, said AWS CEO Andy Jassy in a keynote. AWS has long offered Elastic Container Service (ECS) as its native container orchestration and management offering.
"When we launched ECS, there was no broadly accepted orchestration system for containers," Jassy said. That's changed over the last 18 months as Kubernetes adoption ramped.
Jassy said, despite the previous absence of a hosted service, the majority of Kubernetes implementations running in the cloud have been deployed on AWS. Yet "there's work to do" in easing that experience for customers.
EKS "makes running Kubernetes on top of AWS much, much easier," Jassy said. It can deploy Kubernetes masters across multiple availability zones to avoid single points of failure, and gives users control over upgrading and patching.
Complementing the new container offering is AWS Fargate, which allows users to run either EKS or ECS to deploy containers without having to manage servers or clusters.
"It’s a tricky problem. We worked on that for the last year," Jassy said. "People want to run containers at the task level and not the server level."
Database innovation was another big theme in Jassy's keynote.
"The last 20 years have been a very uncomfortable, unpleasant place with the database providers [enterprises] had to use," he said. Some old-guard database vendors—Oracle the only one specifically mentioned—are "folks who don’t care much about their customers," Jassy said.
"Freedom is the ability not to be locked into abusive or onerous relationships," he added.
Amazon Aurora has done a lot to remedy those issues, delivering a database with scale-out architecture, high-performance and high-availability that comes in at a fraction of the price of its old-guard competition, he said.
That database service has been upgraded with multi-master replication scale out capability for both reads and writes, Jassy said. Aurora now can also scale out across multiple data centers, and AWS will add multi-regional, multi-master capabilities in 2018.
AWS also released in preview Aurora Serverless—an on-demand serverless version of the database that automatically scales, further helping customers save on their spend.
It "automatically scales up when database busy, down when it's not, shuts down when not in use at all," Jassy explained.
Other upgrades came to the Dynamo DB NoSQL database that's often used in mission-critical environments.
Jassy announced the launch of Dynamo DB Global Tables, a fully managed, multi-master, multi-region database; and introduced Dynamo DB Backup and Restore, which continuously backs up data and will soon be able to restore to any point in time over the previous 35 days.
AWS also launched Amazon Neptune, a fully managed graph database.
The variety of database solutions illustrates how much that technology has evolved to become more specialized for different use cases.
"You don’t use relational databases for every application," Jassy said. "That ship has sailed."
On the machine learning front, AWS sees the major challenge to be extending the technology to more users by making it less difficult, less threatening, and more transparent, Jassy said.
"How do we turn machine learning into a capability of a few, to one many more people can take advantage of?" he asked. "The hype and the hope here is tremendous."
When measuring overall use, AWS is the leader in delivering that technology, Jassy said, with twice as many customers using machine learning tools and frameworks on Amazon's cloud than any other provider, Jassy said.
Still, its early days for most customers, especially mainstream enterprises.
"All of them want to be using machine learning," he said.
But there just aren't that many machine learning experts in the world.
"We want everyday developers and scientists to be able to use machine learning much more extensively," he said.
To that end, Jassy introduced Amazon SageMaker, an "easy way to build, train and deploy machine learning models for everyday developers."
The service is modular, allowing users to build models, train algorithms, and host them in different environments.
Amazon's underlying machine learning technology has also been applied to powering a new set of audio and video capabilities.
One is AWS DeepLens, a wireless video camera integrated with deep learning hardware and tools. Developers can build computer-vision models for that device by leveraging SageMaker, and use the AWS GreenGrass IoT platform to set triggers on AWS Lambda serverless compute functions.
Jassy also introduced Amazon Batch Rekognition Video, which delivers real-time video recognition. Amazon Kinesis Video Streams improves secure ingestion and storage of those videos, or any time-encoded data stream.
Other new services are Amazon Transcribe for automatic speech recognition and Amazon Translate for artificially intelligent translation of text between languages. Those products can be used extensively to convert audio to text, then translate it into different languages.
But often those products produce reams of text that would take far too long for anyone to read. To that end, the new Amazon Comprehend service delivers fully-managed, natural language processing to cull through large stores of voice data converted to text.