Amazon re:Invent 2018: 5 key takeaways
by Jim Lundy
Amazon held its annual re:Invent customer and partner conference last week and had record attendance—we’d put it higher than 50,000—and announced over 140 new product features. The growth of AWS, which is led by President Andy Jassy, is what stood out to us, as AWS is a U.S. $27 billion business that’s growing at 45% a year.
This blog highlights the five major takeaways from re:Invent.
1. Machine Learning: The Growth of AWS SageMaker
Amazon threw down a gauntlet on machine learning at re:Invent. SageMaker is only a year old but this year, it received a major upgrade and a number of new offerings.
SageMaker machine learning offerings include:
- SageMaker Neo: train once and get better performance.
- SageMaker RL: Amazon SageMaker Reinforcement Learning (RL) is all about speeding up the training of data sets. SageMaker RL is a managed reinforcement learning service that leverages two third-party algorithms (Intel Coach and Ray ML) and two simulation models (Simulink and MATLAB). It also works with AWS RoboMaker.
- SageMaker Ground Truth: makes it easier to tag data (by using employees, third parties, or a mechanical turk) so the data can then be used to train the ML algorithm.
- Textract: this is the new document analytics capability that can extract text from a document automatically and can do it in volume. Textract is still in preview and we expect AWS to make additional enhancements as it gets rolled out.
- Amazon Rekognition: Amazon’s image and video analytics deep learning algorithm was updated before re:Invent. New capabilities include label detection (the ability to recognize objects in a scene, like a dog or a ball) to allow scene detection within images or videos. Being able to do this automatically will speed up the process of finding the right scene (in a video) or the right image.
AWS Marketplace for Machine Learning
The move that enterprises will really like is that Amazon just introduced one of the first marketplaces for machine learning. The marketplace includes over 150 algorithms and models that can be used with Amazon SageMaker.
Amazon, TensorFlow, and Amazon Chipsets
In a surprising move, AWS is getting into ML chips and announced AWS Inferentia, a machine learning inference chip custom designed by AWS. AWS Inferentia provides hundreds of teraflops per chip and thousands of teraflops per Amazon EC2 instance. It also supports multiple frameworks (e.g., TensorFlow, Apache MXNet, and PyTorch), and multiple data types (e.g., INT8 and mixed precision FP16 and bfloat16).
Amazon also announced Amazon Elastic Compute Cloud (EC2) GPU instances that it refers to as P3dn.24xl. These instances include eight NVIDIA V100 GPUs, 32GB GPU memory, fast NVMe storage, 96 Intel “Skylake” vCPUs, and 100Gbps networking. This move challenges some on-premise providers of deep learning servers, since EC2 can just be turned on.
Based on all of these moves, including building its own chip, Amazon has demonstrated its commitment to machine learning—and this represents a challenge to Google, IBM, and Microsoft. Amazon’s innovation is fueling its growth and its AI/ML announcements made a statement to the market.
2. Database: Challenging Oracle with Aurora
If there were any key point that Andy Jassy wanted to get across, it was that Amazon is now the database alternative to Oracle. Amazon previously announced RDS support for VMware vSphere, which means its Relational Database Service can run in the cloud or on-premise. Amazon’s Aurora Global Database can run across AWS Regions with fast processing times.
There were other announcements including DynamoDB transactions, DynamoDB on-demand, and Amazon Timestream, which is ideal for capturing data and logs from IoT sensors. However, the big news in our opinion was a new ledger-focused database that will leverage the blockchain market.
New Blockchain Ledger Database
The challenge with blockchain is who can you trust and who controls the data?
Amazon’s new QLDB is a ledger database owned by a single trusted entity and shared with organizations that work together. This makes QLDB an immutable and cryptographically verifiable ledger.
On top of QLDB, Amazon enhanced its blockchain offerings with Amazon Managed Blockchain, a service that allows enterprises to set up an Ethereum and Hyperledger Fabric blockchain in a few steps.
3. Hybrid Cloud with Outposts
With AWS Outposts, enterprises can deploy a hybrid cloud option—essentially AWS hardware that runs on-premise. This makes AWS a much bigger player in government, which prefers private cloud options. This also challenges IBM and Oracle—both sell a lot of hardware for data centers.
Aragon has predicted the death of the data center, but a private cloud, which is a managed service, is still going to be a popular option for the next several years.
4. Partner Marketplace Enhancements
The partner marketplace was really one of the big moves at re:Invent. To us, it looked like a strong countermove to slow down Microsoft’s Azure partnering efforts. Nearly anything you want to do on AWS, you can now find partners or consulting organizations that can help you.
Amazon also introduced private pricing, so enterprise software providers can price their own offerings that run on AWS.
5. Deep Racer: A Virtual Development Environment That Stole the Show
AWS DeepRacer offers a way to get your hands on RL, experiment, and learn through autonomous driving. You can get started with the virtual car and tracks in the cloud-based 3D racing simulator, and for a real-world experience, you can deploy your trained models onto AWS DeepRacer and race your friends, or take part in the global AWS DeepRacer League.
Developers: the race is on. The DeepRacer car shown at re:Invent is now backordered until March.
Amazon put on a great show and backed up its own hype with some of the strongest revenue growth of any provider today. To keep that growth going, Amazon threw down the largest set of product announcements in its history. We predict this will ignite a new round of cloud wars, where the focus is on AI and partner marketplaces.
There is more to watch as we head toward 2019. Developing.