I recently joined Chris Presley in an episode of Cloudscape Podcast to share announcements about the latest products and updates that from the 2018 AWS re:Invent conference. Some of the highlights of our discussion included Aurora, Transit Gateway and Timestream and the updates made to DynamoDB and Lambda. We also discussed drone hardware, satellites and text extraction.
The theme of machine learning seemed to run through a lot of the talk at re:Invent. There were a lot of announcements in the field of machine learning technologies and anything AI-related.
I decided to choose what I think are some of the coolest AWS announcements. So, in this blog post, I am going to go over a few details regarding the following announcements:
- AWS Outposts
- AWS DeepRacer
- AWS Satellites
- Machine Learning Services
AWS Outposts was one of the announcements. We don’t have a whole lot of details on how it works yet, but basically, you will be able to run AWS on-prem, which is really cool. You will receive hardware that you can rack yourself and you will administer that hardware from your AWS console transparently.
This comes in two flavors: 1) VMware Cloud on AWS Outposts allows you to use the same VMware control plane and APIs you use to run your infrastructure, and 2) AWS native variant of AWS Outposts allows you to use the same exact APIs and control plane you use to run in the AWS cloud, but on-premises.
AWS will ship out the hardware (you can order single servers, quarter, half or full racks) and it will be up to the customers to follow instructions and physically install the hardware in your co-lo.
I think it’s really cool to actually be able to manage your own hardware, AWS style. It is kind of weird, however, that it contradicts the whole “move to the cloud” movement. This perspective is kind of different, but there are some customers who need this approach, perhaps to adhere to their legal policies or maybe just because they are not ready to move to the cloud.
This looks like a very good intermediate approach to start moving to the cloud or maybe just giving customers the feel for the cloud on-prem.
AWS DeepRacer looked really appealing, like a really cool Christmas toy. It’s a rover: a remote control (toy) car packed with sensors.
The idea is to collect data with the rover and push it to the machine learning services online in AWS. The predictions from the machine learning pipeline can be used for multiple use cases, one of which could be to drive the rover itself. I’m expecting to see a lot of open source code development around these and see people start building some really fancy projects.
It runs on the robotic open source (ROS) toolkit which has been around for a while, so anyone who plays around with these kinds of toys is going to be familiar with them. I am sure we’re going to see this used in school projects and university projects where students are going to learn how machine learning works. Definitely a good play there by AWS.
People might think that AWS Satellites is a little extravagant. However, space launch is becoming cheaper every day. We see it with the success of SpaceX and the reduction of costs to place satellites into orbit. It’s only a matter of time until more companies, not only large ones, are able to send their own satellites into space to create their own data networks. This AWS service is “ground-control” as a service. It would still be too expensive to own your ground control installation, so having this as a service to connect to that mesh of satellites out there and manage it is really cool.
AWS already has some customers using this service: big names in aerospace and defence industries are already customers. I could see this being democratized in the next two years to smaller companies that will invest in their own satellite networks of data.
Machine Learning Services announcements
AWS announced new tools and updates to existing ones, such as SageMaker. We also have new services such as Ground Truth RL, which is reinforcement learning. We have a marketplace for all the machine learning tools and we’re seeing not only specific services but also an ecosystem starting to emerge around the AWS machine learning services. We can see that AWS is tackling the ML-Ops issue.
In the machine learning space, data scientists always have weird sources for their data and it’s not always curated nicely. Sometimes the data comes as dirty as having to scan sheets of paper. Amazon Text Extract is meant for this very reason: it is a very advanced OCR (Optical Character Recognition).
That’s a kind of building block to start playing around with machine learning. You’re going to need a lot of original data and this is a way for you to create that source data that you will input into your machine learning algorithms.
We also have machine learning insights for QuickSight. That’s kind of an outlier detection approach. It’s a service that tracks our latest trends to find out outliers in data and forecast future results. Use cases are to identify business drivers, summarize data in simple ways, make nice presentations and show those transient outliers. There’s a preview of Insights right now for Amazon QuickSight. This simplifies a whole lot the analysis and interpretation of data.
Hear the full conversation which included discussions on many more topics such as AWS Security Hub updates, Transit Gateway, Control Tower. Be sure to subscribe to the podcast to be notified when a new episode has been released.