AWS re:Invent 2019 started on Monday the 2nd of December. We are on day 2 and the keynote by Andy Jassy – CEO of AWS has brought us a whole bunch of new features for existing products but also a couple of new services. Let’s list here all the announcements for future reference:
The new Graviton 2 processor
It’s a custom AWS designed process that is build using a 7 nm (nanometer) manufacturing process. It is based on 64-bit Arm Neoverse cores and it can deliver performances of up to 7x compared to the A1 processor. It must be noted that it has twice the floating point performance of A1 and the additional memory channels and double-sized per-core caches speed memory access by up to 5x. Let’s not go into further details, you can find those on the AWS blog here. All I can say that if I were in charge at Intel I would start paying attention. In fact they should have done that since the last re:Invent.
Availability: Coming soon, limited availability (in preview) of the M6g instances for testing on non-production workloads.
Inf1 Instances with AWS Inferentia Chips
New instances for Machine Learning with AWS Inferentia chips, these were designed to provide fast, low-latency inferencing. Available in four sizes (xlarge, 2xlarge, 6xlarge, 24xlarge), these instances make use of custom Second Generation Intel Xeon Scalable (Cascade Lake) processors.
The Amazon Deep Learning AMIs have been updated and now contain versions of Tensorflow and MxNet optimized for use in Inf1 instances.
For the full details read the article on the AWS blog.
Availability: Available now
Amazon EKS on AWS Fargate now generally available
Amazon Elastic Kubernetes Service allows you to run Kubernetes pods on AWS Fargate. You no longer have to worry about patching, scaling and securing a cluster of EC2 instances to run Kubernetes applications in the cloud. For more details head over to the blog post by Amazon.
Availability: Available now
S3 Access Points
S3 Access points are a new way to manage data access at scale for shared data sets in S3. These are unique hostnames with dedicated access policies that describe how data can be accessed using that endpoint. Access can be either VPC access (no internet) or over the Internet. More details can be found here
Availability: Available now
Redshift update – separately optimize (manage) your compute power and storage
A new family of Redshift instances is being introduced, these are based on Nitro and allow you to separately manage your storage and compute requirements. Most existing Redshift customers will get up to 2x better performance and 2x more storage at the same cost. The new RA3 instances are designed with this new storage model in mind. The new managed storage is based on high-performance SSDs. For more details head over the AWS blog post.
Availability: Available now
AQUA (Advanced Query Accelerator) for Redshift
This is a new query accelerator for Redshift. It will offer 10x better query performance then what is currently available.
Availability: Coming soon (You can sign up for preview here)
Redshift – Data Lake export and Federated Query
Two new features for Redshift:
- Data lake export allows you to unload data from a Redshift cluster to S3 in Apache Parquet format
- Federated query – you can query data across the Redshift cluster, data stored in your S3 data lake and data stored in one or more RDS (PostgreSQL) and Aurora (PostgreSQL) database
Read more about this here
Availability: Available now
UltraWarm for Amazon Elasticsearch
This is a warm storage tier for Amazon Elasticsearch. It’s an extension of the Elasticsearch service by Amazon and offers up to 900TB of storage at a 90% cost reduction over existing alternatives. You can visualize and query your hot and UltraWarm data in a familiar Kibana interface. More informations can be found here
Availability: Preview in US East (N. Virginia, Ohio) and US West (Oregon)
Amazon Managed Apache Cassandra Service
Cassandra was always a pain to manage and naturally customers requested from AWS to do something about that. So now AWS announced Amazon MCS (Managed Cassandra Service), a scalable, highly available and managed Apache Cassandra compatible database services. MCS is serverless so you only pay for the resources you use and the service can automatically scale to accommodate your traffic. MCS provides single digit millisecond read and write performance at any scale and there is no limit on the size of the table or the number of the items. Also you are not required to provision storage. Your data is replicated automatically three times across multiple availability zones. Read more about MCS here.
Availability: Open Preview in US East (N. Virginia, Ohio), Europe (Stockholm), Asia Pacific (Singapore, Tokyo)
Amazon SageMaker Studio
It is a web based IDE (Integrated Development Environment) for Machine Learning. Amazon SageMaker Studio unifies all the tools you need for ML development. You can write code, track experiments, visualize data and perform debugging all within a single interface. Read more about SageMaker Studio here
Availability: Available now in US East (Ohio)
Amazon SageMaker Experiments
This new feature allows you to captures input parameters, configuration and results automatically for your models. The main goal is to make it as simple as possible to create experiments – something that is routinely done during the lifetime of an ML project. Read more about Experiments here.
Availability: Available now in all regions where SageMaker is available
Amazon SageMaker Debugger
A new feature of Amazon SageMaker that is on by default and which automatically identifies complex issues during Machine Learning training jobs. Read more about this feature here.
Availability: Available now in all regions where SageMaker is available
Amazon SageMaker Model Monitor
This new capability of SageMaker is able to automatically monitor and detect concept drifts in deployed models. Usually such detections are difficult, you need to capture the data, run statistical analysis and compare that data to the training set, define rules to detect drift and so on. Model Monitor does this for you. If you are interested in more details head over to the AWS blog post.
Availability: Available now in all regions where SageMaker is available
Amazon SageMaker Autopilot
Autopilot for SageMaker allows automatic training with no loss of visibility or control. With a few clicks in SageMaker Studio you can invoke Autopilot. It will inspect your data set, run a number of candidates to figure out the optimal combination of preprocessing steps, machine learning algorithms and hyperparameters. It uses all this to train an Inference Pipeline which you can easily deploy. Read this article for more informations.
Availability: Available now in most of the AWS Regions.
Amazon SageMaker Processing
SageMaker Processing is a new capability of Amazon SageMaker, it comes with a new Python SDK that lets you easily run your preprocessing, postprocessing and model evaluation workloads on AWS’s managed infrastructure. Read more about SageMaker Processing here.
Availability: Available now in all regions where SageMaker is available
The Deep Graph Library is available on SageMaker
This Python open source library is built for easy implementation of graph neural networks. First released on GitHub in 2018, the library is now available on Amazon SageMaker. Read more about it here.
Availability: Available now
Amazon Fraud Detector
It is a new managed Machine Learning service fraud detection/management.
Availability: In Preview (Sign up for the preview here)
Amazon CodeGuru
CodeGuru is a new machine learning service for automatic code reviews and application profiling. CodeGuru analyzes your pull requests and automatically flags things such as data leaks, concurrency issues and inefficient use of AWS resources. It integrates with GitHub and AWS CodeCommit. The service can also uncover performance bottlenecks with its profiling capability. While it sounds like one of the most promising service announced today don’t rush in unless you are a Java developer. The service only supports Java both for code reviews and for application profiling. Probably more languages will be added later. Still if you want to give it a testdrive head over to the service here.
Availability: Available now
Amazon Contact Lens for Amazon Connect
This is a new machine learning powered contact center analytics service. There are no further details available yet.
Amazon Kendra
Is a highly accurate and easy to use enterprise search that is powered by Machine Learning. You can connect your data sources, create an index and test your search experience on your own content. Kendra uses ML to understand the context of the question. It uses natural language processing and improves automatically over time. The service If you want to try it out head over to the Kendra landing page
Availability: Available now
Outposts is now generally available
Announced last year at re:Invent Outposts is now generally available. Outposts is a hardware offering, basically a mini AWS datacenter that you can order and install into your data center or co-location facility. Read more about it, here.
AWS Local Zones – Los Angeles
The Local Zone is a new type of AWS Infrastructure deployment that brings select AWS services closer to a particular geographic area. This is targeted at customers who need very low latency (single digit milliseconds). The first local zone is in the Los Angeles, California area. In order to use a local zone you need to opt in and crate a new VPC subnet in the Local Zone, this way you can take advantage of all the features that VPC offers. Then you can target the Local Zone when launching EC2 instances and other resources. A very good blog post about this new infrastructure deployment type can be found on the AWS blog.
Availability: Available now
AWS Wavelength
This new offering allows you to build applications that deliver single digit millisecond latencies to mobile devices and users with AWS compute and storage at the edge of 5G network, basically it extends AWS infrastructure to 5G networks. This is a service created in collaboration with Verizon and available only on that network for now. No further details are available yet.
Featured image credit: AWS