Andy Jassy announced big things in his AWS re:Invent 2017 keynote address. With major reveals in the Machine Learning, IoT, Compute, Database, and Networking & Content Delivery service verticals, AWS has pushed their already incredibly high bar even higher. We at 1Strategy are super excited about these new services—our Slack channel has been buzzing all week. Below, I want to share some of the new products and services that we can’t wait to start building with and sharing with our customers.
In the AI and Machine Learning vertical, AWS has announced AWS DeepLens, Amazon Comprehend, Amazon Rekognition Video, Amazon SageMaker, Amazon Translate, and Amazon Transcribe.
AWS DeepLens is a deep learning-enabled wireless HD video camera, not an AWS service. Yeah, you heard that right. A deep learning–enabled video camera. It is fully programmable. As in, you can buy one of these, have it delivered to your house, and start writing code to figure out if your toddler or your dog is the one who keeps triggering the lights in the middle of the night. We’ve already seen smart-home enabled security solutions from big-name security companies and startups, but with limited customizability. DeepLens could take us to previously unseen levels of home security automation.
With Amazon Comprehend, AWS adds scalable, online, continuously improving natural language processing (NLP) capabilities to its portfolio. As of today, it provides APIs for Sentiment Analysis, Entity Recognition, Language Detection, Topic Modeling, and Language Detection. With integration to S3 and Glue, you’re able to build scalable ETL and analytics pipelines with Comprehend’s insights as the source.
Amazon Rekognition Video
Amazon Rekognition, AWS’s deep learning-based image analysis service, has added support for video streaming with Amazon Rekognition Video. This service has object recognition and labelling functionality that enable computer vision tasks on live streams.
Amazon SageMaker is a brand new, fully-managed service that enables scalable building, training, and deploying of machine learning models. SageMaker aims to make the life of developers and data scientists easier by removing many of the barriers that typically impede machine learning at scale, such as paralleling the training task, dynamically scaling to meet load requirements, and dealing with the infrastructure components that lift models into production. SageMaker has exciting applications in many of the traditional market segments that are already leveraging machine learning, such as AdTech, financial modelling, language processing, and security. Especially exciting are the possible applications of SageMaker in IoT, in conjunction with the other fantastic machine learning and AI services released by AWS this year (think Amazon DeepLens and Amazon Rekognition Video). A future where we can program against our own, custom built, and deep learning-enabled home security systems is here.
With Amazon Translate, AWS has entered the natural language translation space in a big way. Translate, a neural machine translation service, leverages deep learning to provide real-time automation of language translation. By leveraging this deep learning approach, AWS now boasts a scalable, performant, and highly-accurate solution to meet the language translation needs of their customers. With Translate, customers will now be able to automate their language localization processes at whatever scale they require.
As its name suggests, Amazon Transcribe is an automated speech recognition (ASR) service that enables customers to add speech-to-text capabilities to their applications, analyze large-scale collections of audio, and efficiently produce accurate text files that contain transcriptions of this text. Transcribe is a managed service that continually learns and adapts to changes in language structure and use in order to provide accurate transcriptions. Translate currently supports conversion of Spanish and English, with support for more languages coming in the (near) future.
In the IoT vertical, they’ve announced AWS IoT Device Management, AWS IoT Analytics, AWS IoT 1-Click, AWS IoT Device Defender, Amazon FreeRTOS, and AWS Greengrass ML Inference. As you’ll see as you read on, AWS has invested heavily in IoT this year, with each of their services promising well-engineered solutions to hard problems.
AWS IoT Device Management
The new AWS IoT Device Management service acts as a control hub for IoT devices, providing functionality for device registration (both individually and in bulk), permission scoping and authorization of devices, organization of devices into groups, monitoring, and updating. If you have fleets of IoT devices that need to be managed at scale, AWS IoT Device Management offers solutions for you.
AWS IoT Analytics
Collecting, processing, and analyzing the massive volumes of unstructured data produced by fleets of IoT devices is a challenging problem in both complexity and scale. AWS IoT Analytics offloads all of these challenges by offering a completely managed and automatically-scaling IoT analytics platform that handles ingestion, enrichment, storage, and analysis. Even more exciting, IoT Analytics provides tooling that enables machine learning against your IoT data. By connecting your IoT data directly to hosted Jupyter notebooks, you can design, train, and deploy machine learning models at scale.
AWS IoT 1-Click
With AWS IoT 1-Click, you can provision single-click devices that come preconfigured to execute your custom lambda code in the cloud. This service allows you to map individual lambda functions to single devices or groups of devices. IoT 1-Click connects to AWS over Wi-Fi or cellular network (depending on which device you purchase). You can be sure to expect that AWS will be adding more supported devices to the list of options soon.
AWS IoT Device Defender
With the emergence of cloud-based IoT integration, ensuring the security of your fleets of IoT devices is more important than ever. AWS IoT Device Defender makes doing so much easier by continually auditing the security policies you implement for your devices, in order to check if, when, and where they deviate from security best practices. Even better, IoT Device Defender will send out alerts when such deviations occur, so you are always aware of the security posture of your fleet. This event-based alert functionality, as we’ve come to expect from AWS, comes with integration with other AWS services.
Based on the FreeRTOS microcontroller kernel, Amazon FreeRTOS extends the base kernel distribution with Amazon-specific software, making it easy to integrate devices with AWS services in the cloud. If you took the familiar Amazon Linux operating system popular for instance deployments in EC2, blasted it with a shrink-ray, and optimized it for small, IoT devices, you’d effectively have Amazon FreeRTOS.
AWS Greengrass ML Inference
AWS Greengrass ML Inference is part of AWS’ heavy investment this year in technology operated at the edge (i.e., deployed onto devices and network-peripheral systems, in order to achieve the related performance and data-locality benefits). With Greengrass ML Inference, you can now develop and train your machine learning models in the cloud and then deploy onto devices that are able to connect back to the cloud. Want to train a model that knows how to identify cats, specifically your cat, which uses the power of big data on the cloud, but utilizes a camera-enabled device in your home? With Greengrass ML Inference, you can do this.
In the Compute vertical, AWS Fargate and Amazon ECS for Kubernetes (EKS) have been announced.
With Fargate, AWS adds something pretty slick on top of current container orchestration and deployment platforms: the ability to deploy containers without having to worry about servers and clusters at all. With both ECS and Kubernetes, individually, even though the deployment problem is pretty much solved, administrators still have to be concerned with the server clusters underlying their systems.
With Fargate, AWS customers looking to deploy and orchestrate container clusters are given two modes: the Fargate launch type and the EC2 launch type. The Fargate launch type restricts the configurability of the underlying servers for your application, handling the underlying infrastructure for you. The EC2 launch type gives you more configuration leeway by providing access to the EC2 instances underlying your container clusters. AWS has given us two great options—all that’s left is for us to decide which one best suits our needs. Guidance regarding how to make this choice may be the subject of a future blog post.
Amazon ECS for Kubernetes
AWS customers have been asking for this for a long time. There isn’t much more to say than we are excited it’s finally here. Now that AWS offers a fully-managed Kubernetes service, we no longer have to worry about the pain of managing custom deployments on our EC2 nodes. With full compatibility with existing open-source Kubernetes environments and a commitment to running the up-stream version of open-source Kubernetes, migrating away from our EC2-based deployments ought to be relatively painless, insofar as orchestration is concerned. Thank you, AWS!
In the Database vertical, Amazon Neptune, Amazon Aurora Multi-Master, Amazon DynamoDB On-Demand Backup, and Amazon DynamoDB Global Tables were announced.
AWS Neptune is AWS’ first, and fully-managed graph database service. Focusing on high performance and reliability, Neptune is optimized for storing billions of relationships between the nodes of your graphs. What’s more, Neptune provides millisecond-speed query functionality on top of its great scalability. Even better, Neptune provides many of the great features we’ve come to expect from RDS, including high-availability, read-replicas, point-in-time-recovery, and continuous, automated backups. Neptune supports both the Gremlin and SPARQL graph query languages.
Amazon Aurora Multi-Master
With the release of Amazon Aurora Multi-Master, you can now create multiple read/write database instances across multiple availability zones. This is a big deal! Previously, Aurora databases were restricted to a single master node and a collection of read-only replica instances, with the consequence of limited scalability for writes. With Multi-Master, you can now scale out your write workload. By providing distribution of multi-master instances across availability zones, zero-downtime during failure recovery is now possible. Very, very exciting.
Amazon DynamoDB On-Demand Backup
DynamoDB just got a whole lot better, with the addition of on-demand backup features. Now you can create full table backups for long-term retention, recovery, and archival storage to meet your governance and compliance needs. As we’ve come to expect with AWS database services, these backups come with no penalty to read-write performance or availability. Backups persist past the deletion of tables, so rest-assured you now have recourse if a table gets dropped.
Amazon DynamoDB Global Tables
One of the most exciting announcements of re:Invent, in my opinion, is Global Tables for DynamoDB. With the addition of this feature, you can now deploy multi-master and multi-region DynamoDB tables, achieving what is nearly a holy grail for availability and reliability. Instead of having to manage an individual DynamoDB table per region, you only have to manage a single table. Granted, there are still situations where you may prefer to deploy the same database multiple times, across regions in order to meet data locality and data sovereignty needs, you now have a much easier option for everything else.
Networking & Content Delivery
In the Networking and Content Delivery vertical, the major news is AWS PrivateLink.
With Amazon PrivateLink, you can now connect supported AWS services, services hosted in other accounts, and supported AWS Marketplace services in your private VPCs. PrivateLink is highly-available, scalable, and doesn’t require you to configure internet gateways, NAT devices, public IP addresses, Direct Connect, or VPN connections in order to communicate with the service. What’s more, PrivateLink allows you to roll your own custom endpoint services and link them together via AWS’ network backbone, never leaving the Amazon network—security, availability, and extensibility to custom services. We love it.
Overall, there were lots of big service release announcements at re:Invent this year. We are looking forward to helping our clients better meet their AWS needs by leveraging these new services, as well as playing around with them ourselves. I, for one, will be using the new machine learning and IoT functionality to tinker with smart home integrations.
What are you looking forward to using these new services and features to accomplish? We’d love to hear your ideas and even help! You can get in touch by emailing me at firstname.lastname@example.org
About the Author
Sophia Hudson is a DevOps Engineer at 1Strategy, an Amazon Web Services (AWS) Consulting Partner. Her domain expertise spans distributed systems, full-stack software engineering, automation, data science, and scientific computing. Sophia works with customers to better meet their goals on AWS. Previously, she worked for Motorola Solutions, Zions Bancorporation, and the Scientific Computing and Imaging Institute.
Sophia holds Bachelor of Science degrees in both Mathematics and Economics, as well as the AWS Certified Developer – Associate and AWS Certified Solutions Architect – Associate credentials from AWS. When she’s not working, you can find her playing electric guitar or watching anime with her cat.