AWS News Blog

New – EC2 M6g Instances, powered by AWS Graviton2

Voiced by Polly

Starting today, you can use our first 6th generation Amazon Elastic Compute Cloud (Amazon EC2) General Purpose instance: the M6g. The “g” stands for “Graviton2“, our next generation Arm-based chip designed by AWS (and Annapurna Labs, an Amazon company), utilizing 64-bit Arm Neoverse N1 cores.

Graviton 2 chipset

These processors support 256-bit, always-on, DRAM encryption. They also include dual SIMD units to double the floating point performance versus the first generation Graviton, and they support int8/fp16 instructions to accelerate machine learning inference workloads. You can read this full review published by AnandTech for in-depth details.

The M6g instances are available in 8 sizes with 1, 2, 4, 8, 16, 32, 48, and 64 vCPUs, or as bare metal instances. They support configurations with up to 256 GiB of memory, 25 Gbps of network performance, and 19 Gbps of EBS bandwidth. These instances are powered by AWS Nitro System, a combination of dedicated hardware and a lightweight hypervisor.

For those of you running typical open-source application stacks, generally deployed on x86-64 architectures, migrating to Graviton2-based instances will give you up to 40% improvement on cost-performance ratio, compared to similar-sized M5 instances. M6g instances are well-suited for workloads such as application servers, gaming servers, mid-size databases, caching fleets, web tier and the likes.

We ran an extensive preview program to collect customer feedback on this 6th generation instance type. For example, Honeycomb uses 30% fewer instances vs C5, KeyDB observes 65% better performance and 20% cost reduction vs M5, InterSystems reported 28% performance improvement and 20% cost reduction compared to equivalent M5 instances, and Treasure Data benchmarked 30% increase of performance for 20% cost reduction compared to similarly sized M5 instances. You can read more customer stories including Hotelbeds, Redbox, Nielsen, Mobiuspace, RayGun on the M6g web page.

Several AWS services teams are evaluating these instances too. For example, during their testing, the Amazon ElastiCache service team found that M6g instances deliver up to 50% throughput improvement over M5 instances on Redis.

Major Linux distributions are available on Arm architecture, just select the Amazon Machine Image (AMI) corresponding to the Arm version of your favorite distribution when launching an instance in the AWS Management Console. Be sure to select the 64-bit (Arm) button on the right part of the screen.

Launch ARM AMI in the console

If you choose the AWS Command Line Interface (AWS CLI) instead, use the corresponding image-id for your region, architecture, and distribution. For example, to start an Amazon Linux 2 instance:

AMI_ID=$(aws ssm get-parameters-by-path --path /aws/service/ami-amazon-linux-latest --output text --query "Parameters[?contains(Name, 'ami-hvm-arm64')].Value")
aws ec2 run-instances --image-id $AMI_ID --instance-type m6g.large --key-name my-ssh-key-name --security-group-ids sg-1234567

(you need to adjust the ssh key name and the security group ID in the above command)

Once the instance is started, it behaves like any Amazon Elastic Compute Cloud (EC2) instance:

~ % ssh ec2-user@ec2-01-01-01-01.compute-1.amazonaws.com
Warning: Permanently added 'ec2-01-01-01-01.compute-1.amazonaws.com,01.01.01.01' (ECDSA) to the list of known hosts.
Last login: Wed Apr 22 12:26:44 2020 from 01-01-01-01.amazon.com

       __|  __|_  )
       _|  (     /   Amazon Linux 2 AMI
      ___|\___|___|

https://aws.amazon.com/amazon-linux-2/
[ec2-user@ip-172-31-16-155 ~]$ uname -a
Linux ip-172-31-16-155.ec2.internal 4.14.171-136.231.amzn2.aarch64 #1 SMP Thu Feb 27 20:25:45 UTC 2020 aarch64 aarch64 aarch64 GNU/Linux

The Arm software ecosystem is broad and deep, from Linux distributions (Amazon Linux 2, Ubuntu, Red Hat Enterprise Linux, SUSE Linux Enterprise Server, Fedora, Debian, FreeBSD), to language runtimes (Java with Amazon Corretto , NodeJS, Python, Go,…), container services (Docker, Amazon Elastic Container Service (Amazon ECS), Amazon Elastic Kubernetes Service (Amazon EKS), Amazon Elastic Container Registry (Amazon ECR)), agents (Amazon CloudWatch, AWS Systems Manager, Amazon Inspector), developer tools (AWS Code Suite, Jenkins, GitLab, Chef, Drone.io, Travis CI), and security & monitoring solutions (such as Datadog, Crowdstrike, Qualys, Rapid7, Tenable, or Honeycomb.io).

You will find Arm versions of commonly used software packages available for installation through the same mechanisms that you currently use (yum, apt-get, pip, npm …). While some applications may require re-compilation, the vast majority of applications that are based on interpreted languages (such as Java, NodeJS, Python, Go) should run unmodified on M6g instances. In the rare cases where you will need to recompile or debug code, we have assembled some resources to help you to get started.

We are not going to stop at general purposes M6g instances, compute optimized C6g instances and memory optimized R6g instances are coming soon, stay tuned.

Now it’s your turn to give it a try in one the following AWS Regions : US East (N. Virginia), US East (Ohio), US West (Oregon), Europe (Ireland), Europe (Frankfurt), and Asia Pacific (Tokyo), or watch to learn about our other recent Graviton2-based instances.

As usual, let us know your feedback.

-- seb
Sébastien Stormacq

Sébastien Stormacq

Seb has been writing code since he first touched a Commodore 64 in the mid-eighties. He inspires builders to unlock the value of the AWS cloud, using his secret blend of passion, enthusiasm, customer advocacy, curiosity and creativity. His interests are software architecture, developer tools and mobile computing. If you want to sell him something, be sure it has an API. Follow him on Twitter @sebsto.