There has been a heap of super exciting announcements at re:Invent today. It was really hard to pick only a few amongst a plethora of amazing new services and enhancements but here are our top picks.
We knew this was coming. AWS is the most common used platform for Kubernetes container-based deployments. Before today, this was entirely self-managed on top of EC2 instances. Now you can take advantage of a fully managed Kubernetes service in the form of EKS, a parallel service to the Docker based ECS.
With this new service, much like ECS, AWS will manage the Kubernetes platform with customers managing the EC2 based worker nodes. It will natively support Kubernetes containers without change and will support a defined set of add-ons. You will also be able to install any add-on you require on the worker nodes.
The ECS platform hasn’t been forgotten at re:Invent and AWS continues to evolve and improve the service. Do you love ECS but don’t want to manage the underlying EC2 instances that make up the worker nodes? We certainly don’t want to deal with scaling the instances! Like us, you need Fargate!
Fargate extends ECS to fully manage your cluster including the worker nodes. No more management of the EC2 instances and having to worry about scale out and scale in. It just works. It will initially be available for ECS but will soon be available for EKS as well.
The story doesn’t end there. ECS is also getting support for Windows based containers. Nicely done, AWS!
GuardDuty is a threat analysis service that integrates with a bunch of AWS services to provide an overall view of the security of your AWS environment. It uses known threat intelligence data as well as unusual activity within your account to provide a security health view of your AWS environment as well as near real-time alerting based on rules that you can customise. It can even report on potentially compromised EC2 instances through suspicious outbound traffic requests. GuardDuty uses the power and insights of the entire AWS Cloud to provide far greater insight than could be gleaned by an individual organisation.
GuardGuty will be an important service for any security professional and, since it can generate events in CloudWatch will be easily integrated into external log management solutions such as SumoLogic and workflow tools like Jira. In fact Sumologic has already announced native integration.
Machine Learning & AI
We are not sure where to begin. There is so much amazing new stuff coming out at re:Invent! Here’s some highlights to whet your appetite.
There’s language-based services.
- Transcribe is a long form version of the existing Lex service designed to transcribe voice files to text. It has the ability to distinguish between multiple speakers and supports multiple languages.
- Translate does what it says on the tin and will initially support 12 languages. It runs in real-time and batch.
- Comprehend is intended to derive entities, meaning, key phrases and sentiment from the input.
Rekognition has been enhanced to support video in both batch and real-time. It included facial recognition and person tracking. Kinesis has been extended to include video streaming support at the same time. CCTV as a service anyone? Just bring a box of cameras.
How’s your machine learning skills? Yep, we thought so. They are pretty hard to come by. If only someone could take away the knitting around model development, training and optimisation so that your developers could start to take advantage of the capabilities it offers. Step forward SageMaker. SageMaker is designed to simplify what is a very complicated process that requires rare and expensive resources to understand it end to end. It makes it easier to manage notebooks, offering pre-canned algorithms, frameworks and associated drivers to get started quickly. The resulting models can then be trained and optimised within SageMaker before executing on the AWS Cloud or exported to be executed somewhere else, should you need to do that.
AWS has also realised that many developers would like to use machine learning but have little experience. They have also released DeepLens, a wireless camera designed for deep learning. It comes with a heap of examples to get started quickly. We are looking forward to getting our hands on one here at re:Invent and building some sample applications.
It’s another great example of the democratisation of technology that AWS provided, lowering the bar of entry into what can be the intimidating world of machine learning and deep learning.
A few years ago, AWS released Aurora, a fully managed MySQL and PostgreSQL compatible database. It’s been widely successful, AWS’s fastest growing service ever, in fact. It offers Enterprise grade relational database performance and features at a 10th the cost of the legacy providers (yes, I’m looking at you Oracle).
Aurora has always supported high availability through read replicas. AWS has now extended Aurora to provide multi-master, massively scaling write performance and effectively meaning zero downtime in the event of a failure of the master node – previously the failure of a master would result in a 30 second or so outage while the cluster promoted a read replica to write capability.
On top of that, Aurora is now available in serverless mode, auto scaling and spinning up and down on demand. It’s perfect for variable workloads that are subject to rapid change.
With these enhancements to Aurora, there’s never been a better time to migrate off legacy and into the AWS Cloud for your database workloads.
Stay tuned for more from re:Invent tomorrow and follow the announcements at https://aws.amazon.com/blogs/.