Wednesday - Day 3 was the first big keynote with Andy Jassy - AWS CEO - and there was a ton of announcements!
Machine Learning
Amazon Forecast
Time series forecasting usually means predicting the next data points of the series, not the future. Even if simple to explain, it’s actually rather complicated to implement correctly.
Amazon Forecast allows you to build your own time series forecasting using the AWS Console, the API or the new SDK.
Amazon Forecast – Time Series Forecasting Made Easy
A Machine Learning marketplace
Using this new service, AWS allows data scientists to share - and sell - their algorithms and models very easily.
For Machine Learning users, they can have access to a very broad offering of models.
Machine Learning algorithms and model packages now available in AWS Marketplace
Amazon Personalize
With Personalize, AWS is adding the service that made Amazon a different e-commerce website: recommendations.
The service is easy to use and allows to ingest user data, products catalog and external data to generate user recommendations.
Real-Time Personalization and Recommendation for Everyone
Amazon SageMaker Neo
SageMaker Neo allows you to execute your Machine Learning models (inference) at the edge for optimal performance.
This could be useful when using the models offline.
Train Your Machine Learning Models Once, Run Them Anywhere
Amazon SageMaker Ground Truth
Anyone who has done machine learning will tell you this, creating a training dataset of good quality and with correct labels is difficult and time consuming.
Ground Truth allows to automate the labeling of your dataset when possible and you will be able to delegate to humans (Mechanical Turk) when the dataset is too complicated to be automatically labeled.
Amazon Sagemaker RL (Reinforcement Learning)
The idea of reinforcement learning is to train an algorithm without the need to have a training dataset. When the algorithm takes a good decision, the agent receives a reward.
After several iterations, the algorithm becomes more and more accurate.
Sagemaker RL completes the existing offer of supervised and unsupervised machine algorithms.
Amazon SageMaker RL – Managed Reinforcement Learning with Amazon SageMaker
AWS DeepRacer
Definitely the funniest announcement of re:Invent so far!
DeepRacer is a 1/18th scale race car designed to be an autonomous vehicle: cameras, sensors, gyroscope... The only missing thing? The best machine learning algorithm to drive the car!
The goal is to open the SDK and track simulator to let developers all around the world create this intelligence. Once your algorithm is ready, you can compete against others in the league AWS has created. There will be events organized all year long, and the winner of each event will be invited to re:Invent 2019.
AWS DeepRacer – Go Hands-On with Reinforcement Learning at re:Invent
Pre-order on Amazon
Database
Amazon DynamoDB On-Demand
Until now, you had to configure a specific throughput (IOPS, I/O per second) when setting up your DynamoDB tables.
AWS now offers to adapt the performance of the table based on the concurrent workload they receive. The billing also auto-adjusts based on the number of requests.
Amazon DynamoDB On-Demand – No Capacity Planning and Pay-Per-Request Pricing
Amazon Timestream - a Time-series database
A time-series database was the last missing piece in the AWS noSQL offer, and this is not the case anymore with Amazon Timestream. A time-series database is designed to ingest the stream of data generated by systems like IoT devices.
Timestream supports the ingestion of data at high throughput, but also analytics and processing specifics to time-series.
Storage
S3 Glacier Deep Archive
This new storage class for Glacier is intended to replace the tapes that companies traditionally use for their long-term storage.
Coming Soon – S3 Glacier Deep Archive for Long-Term Data Retention
AWS Lake Formation
Building your own data lake on-premise can take months. Building it using all the AWS services is faster, but still complicated when trying to integrate all the different services, security, and data governance.
AWS Lake Formation aims at doing this in a matter of days by setting up all the services and configuration for you. Including security management, access management, and data sharing.
Blockchain
AWS Managed Blockchain
Finally! A blockchain service managed by AWS. In a few clicks, you can deploy a blockchain network based on the open source Hyperledger Fabric or the Ethereum network.
Amazon QLDB
When talking about blockchain, the properties that the users are looking for are data lineage and immutability. Amazon Quantum Ledger Database (QLDB) is a new database that provides these guarantees by making its commit log immutable and cryptographically signed.
Amazon Quantum Ledger Database (QLDB)
Accounts management
AWS Control Tower
Drive and manage your cloud migration with AWS Control Tower.
Since earlier this year, we see more and more customers using AWS Landing Zone to offer self-service to their teams guided by templates and best practices. Using Control Tower, you can now reuse and manage templates from AWS catalog in few clicks. You can easily implement rules and compliance checks using this new tool.
Security
AWS Security Hub
When talking about security, it is often difficult to aggregate data points and metrics from all the AWS services across multiple accounts. This often requires to use a third-party tool. AWS Security Hub is now a central location to aggregate and configure alerts for your compliance checks.
Hybrid Cloud
AWS Outposts
VMware CEO Pat Gelsinger was on stage with Andy Jassy to announce the partnership to bring AWS compute power to your own data center with AWS Outposts. A consistent hybrid experience is driving this new hybrid offer and will allow you to keep your most sensitive data in your data center while still taking benefits of the AWS experience.
Comments