Last Friday, Amazon held an online conference called AWS Re:Cap to summarise all the new Amazon Web Services features presented at the end of 2020. Let’s take a look at the exciting new features!
EC2 Compute Announcement
New EC2 P4d Instances
Amazon started with announcing the new EC2 P4d instances, which can support up to 400Gbps networking speed on the cloud. These are the first instances with such a blazing fast networking speed!
But that’s not all, Amazon also announced the new P4d instances to have up to 2.5x better deep learning performance, which could mean up to 60% cost savings in using these fast and powerful instances.
For SAP use cases, Amazon now has a newer largest high memory of 24 TB.
New EC2 M5zn Instances
Further on, new EC2 M5zn instances were discussed, which provide higher processing capabilities. This is especially interesting in the case of software licensing. If you are only allowed to run the software on a specific amount of servers, this could help you save money, as you get more processing power on each M5zn service. These instances offer up to 45% better price-performance compared to the previous EC2 M5 instances.
New EC2 R5b Instances
The new EC2 R5B instances provide increased Elastic Block Store — also called EBS — performance. Technically, providing up to 7500 MB/S bandwidth and up to 260K IOPS of EBS performance. These instances are said to deliver up to 3x better price-performance compared to the previous EC2 R5 instances. These new instances are ideal for larger databases and analytics workloads.
New EC2 G4d Instances
The new G4d instances provide up to 45% better price-performance compared to the previous G4dn instances. This is now the best price-performance for graphic workloads in the cloud, offered by AWS.
New EC2 Mac1 Instances
What? Mac on AWS? Yes, you heard it right! AWS now has EC2 Mac1 instances. Why is this so great? Not every developer has a Mac machine available for use. When they need to work some time on Mac projects, this would enable them to work on a dedicated Mac in the cloud, without having to purchase separate hardware for this.
However, this is not yet perfect. EC2 Mac1 can only be set up for dedicated EC2 instances, meaning the price will also be higher. In fact, one instance in the eu-west-1 Ireland region costs about 1.2 USD per hour (!). Furthermore, Amazon requires a minimum of 24 hours usage for these instances, meaning one day would cost slightly lower than 30 USD.
Please note that On-Demand Mac1 Dedicated Hosts have a minimum host allocation and billing duration of 24 hours. - AWS
While the high cost is somewhat disappointing, it is still a very nice introduction of Mac into the AWS range of services. I’m hoping to see price declines as these instances become more popular.
Amazon announced a new service called AWS Trainium, which is dubbed, the first machine learning chip for training. The most cost-effective in the cloud.
I’m not a machine learning expert, but in summary, the TFLOPS compute power will be higher than other learning instances in the cloud. It will support common machine learning frameworks such as TensorFlow, MXNet, and PyTorch.
AWS Trainium will be available as EC2 instances, but also via SageMaker.
Ideal for image recognition, speech recognition, natural language processing, forecasting, and fraud detection. — AWS
New EBS GP3 Volumes
The new GP3 volumes now offer up to 20% lower price per GB compared to the previous GP2 volumes. Furthermore, GP3 has a baseline of 3000 IOPS, while GP2 only had 3000 IOPS for burst mode. The speeds are much faster as well, up to 1000MiB/s for GP3 compared to up to 250MiB/s for GP2.
Containers: ECS/EKS Anywhere
Amazon announced ECS/EKS anywhere, meaning you could run ECS/EKS containers on-premise. Amazon provides the container orchestration and takes the workload off for operating the clusters on-premise. This is planned to be released this year.
Open-Source EKS Distro
Amazon is working on making the EKS distro open-source, providing you with a well-maintained distro. This distro will have the same testing and security patching cycles as the regular EKS services, which makes this very attractive. In terms of licensing, this will be available under the Apache 2.0 license.
Package and Deploy Lambdas with Containers
Serverless with containers? Yes! The code and dependencies can be packaged using container images, now also providing all the containerization advantages for serverless applications. This brings a consistent way of working with serverless applications. Furthermore, the lambdas could also be deployed on top of third-party base images.
Serverless Applications and Lambda
Serverless 1ms Billing Granularity
This update is more in terms of billing, AWS now allows serverless applications to be billed for each millisecond used for processing requests. This will give development teams another boost and reason to optimize their applications to be fast, in turn also reducing the cost! This is a very nice move from AWS. Have you checked the processing time your serverless applications use up recently?
This change has started since the December 2020 billing cycle.
Aurora Serverless v2 Preview
Even faster scaling and provisioning of databases for applications. It is an improved version of Aurora Serverless v1. The AWS team mentions up to 90% cost savings when under peak load.
Babelfish for Amazon Aurora PostgreSQL
A common issue is the high costs associated with running MS SQL on AWS, mainly due to Microsoft licensing terms. Amazon created a new service called Babelfish to reduce this cost. Babelfish is a layer they have developed on top of PostgreSQL which can understand T-SQL queries.
This will enable development teams to complete migrations to AWS faster since there would be a more affordable and acceptable option to host MS SQL databases on AWS.
Babelfish for All
Amazon will open-source Babelfish in 2021. It will be available on Github and would allow anybody to run and use the Babelfish T-SQL translation layer for PostgreSQL databases. It will be available under the Apache 2.0 license.
Amazon Lookout for Metrics
Amazon Lookout will enable you to detect anomalies in metrics faster, as well as find the underlying root causes. Amazon is doing this by machine learning their 20+ years of experience.
Redshift Machine Learning
Redshift will enable users to work with Machine Learning, without needing to learn new technologies or tools.
Amazon Neptune ML
Neptune ML is a new capability on Neptune, allowing for running machine learning functionalities on fully managed graph databases.
It can work with Graph Neural Networks, which can be much more accurate in terms of predictions.
Last but not least, Amazon announced AWS CloudShell, a full AWS CLI available from the browser. CloudShell automatically saves your files and settings between shell sessions.
Amazon has released exciting improvements and services. Mac1 EC2 instances, generally improved performances, Babelfish, open-sourcing EKS distro, billing serverless by the millisecond, and so on.
What are you the most excited about? What updates would you like to see in 2021? Let us know in the comments!