When to Use AWS Graviton2 Instances

Advantages of Graviton Processor Instances for Your Workloads

Amazon Web Services (AWS) offers Elastic Compute Cloud (EC2) Graviton2 processor-based instances utilizing 64-bit Arm Neoverse cores.

AWS Graviton2 processors deliver a significant leap in performance and capabilities over the first-generation AWS Graviton processor (A1 series) instances. Amazon boasts that these new instances provide up to a 40% better price and performance over comparable current generation x86-based instances for a wide variety of workloads.

Amazon has released Graviton2 instances across General Purpose, Compute Optimized, and Memory Optimized families in M6g, C6g, and R6g types. For applications that require local storage, NVMe SSD local instance storage variants like C6gd, M6gd, and R6gd are available.

General Purpose AWS EC2 Graviton2 Instances
Instance SizevCPUsMemory (GiB)Instance Storage (GIB)Network Bandwidth (Gbps)EBS Bandwidth (Mbps)
m6g.medium14EBS-onlyUp to 10Up to 4,750
m6g.large28EBS-onlyUp to 10Up to 4,750
m6g.xlarge416EBS-onlyUp to 10Up to 4,750
m6g.2xlarge832EBS-onlyUp to 10Up to 4,750
m6g.4xlarge1664EBS-onlyUp to 104,750
m6g.8xlarge32128EBS-only129,000
m6g.12xlarge48192EBS-only2013,500
m6g.16xlarge64256EBS-only2519,000
m6g.metal64256EBS-only2519,000
m6gd.medium141x59 NVMe SSDUp to 10Up to 4,750
m6gd.large281x118 NVMe SSDUp to 10Up to 4,750
m6gd.xlarge4161x237 NVMe SSDUp to 10Up to 4,750
m6gd.2xlarge8321x474 NVMe SSDUp to 10Up to 4,750
m6gd.4xlarge16641x950 NVMe SSDUp to 104,750
m6gd.8xlarge321281x1900 NVMe SSD129,000
m6gd.12xlarge481922x1425 NVMe SSD2013,500
m6gd.16xlarge642562x1900 NVMe SSD2519,000
m6gd.metal642562x1900 NVMe SSD2519,000
Compute Optimized AWS EC2 Graviton2 Instances
Instance SizevCPUsMemory (GiB)Instance Storage (GiB)Network Bandwidth (Gbps)EBS Bandwidth (Mbps)
c6g.medium12EBS-onlyUp to 10Up to 4,750
c6g.large24EBS-onlyUp to 10Up to 4,750
c6g.xlarge48EBS-onlyUp to 10Up to 4,750
c6g.2xlarge816EBS-onlyUp to 10Up to 4,750
c6g.4xlarge1632EBS-onlyUp to 104750
c6g.8xlarge3264EBS-only129000
c6g.12xlarge4896EBS-only2013500
c6g.16xlarge64128EBS-only2519000
c6g.metal64128EBS-only2519000
c6gd.medium121x59 NVMe SSDUp to 10Up to 4,750
c6gd.large241x118 NVMe SSDUp to 10Up to 4,750
c6gd.xlarge481x237 NVMe SSDUp to 10Up to 4,750
c6gd.2xlarge8161x474 NVMe SSDUp to 10Up to 4,750
c6gd.4xlarge16321x950 NVMe SSDUp to 104,750
c6gd.8xlarge32641x1900 NVMe SSD129,000
c6gd.12xlarge48962x1425 NVMe SSD2013,500
c6gd.16xlarge641282x1900 NVMe SSD2519,000
c6gd.metal641282x1900 NVMe SSD2519,000
Memory Optimized AWS EC2 Graviton2 Instances
Instance Size vCPUs Memory (GiB)Instance Storage Network Bandwidth (Gbps) EBS Bandwidth (Mbps)
r6g.medium18EBS-onlyUp to 10Up to 4,750
r6g.large216EBS-onlyUp to 10Up to 4,750
r6g.xlarge432EBS-onlyUp to 10Up to 4,750
r6g.2xlarge864EBS-onlyUp to 10Up to 4,750
r6g.4xlarge16128EBS-onlyUp to 104750
r6g.8xlarge32256EBS-only129000
r6g.12xlarge48384EBS-only2013500
r6g.16xlarge64512EBS-only2519000
r6g.metal64512EBS-only2519000
r6gd.medium181x59 NVMe SSDUp to 10Up to 4,750
r6gd.large2161x118 NVMe SSDUp to 10Up to 4,750
r6gd.xlarge4321x237 NVMe SSDUp to 10Up to 4,750
r6gd.2xlarge8641x474 NVMe SSDUp to 10Up to 4,750
r6gd.4xlarge161281x950 NVMe SSDUp to 104,750
r6gd.8xlarge322561x1900 NVMe SSD129,000
r6gd.12xlarge483842x1425 NVMe SSD2013,500
r6gd.16xlarge645122x1900 NVMe SSD2519,000
r6gd.metal645122x1900 NVMe SSD2519,000

AWS initially offered the M6g instances in 8 sizes:

Instance NamevCPUsMemory (GiB)Price/Hour (Linux)Price/Hour (RHEL)
m6g.medium14$0.0385$0.0985
m6g.large28$0.0770$0.1370
m6g.xlarge416$0.1540$0.2140
m6g.2xlarge832$0.3080$0.4380
m6g.4xlarge1664$0.6160$0.7460
m6g.8xlarge32128$1.2320$1.3620
m6g.12xlarge48192$1.8480$1.9780
m6g.16xlarge64256$2.4640$2.5940

Considerations for Migrating Workloads to Amazon EC2 Graviton2 Instances

The new Graviton2 instances are supported by several open-source software distributions and services.

Here is the list of support for Graviton2 instances:

  • Amazon Linux 2, Ubuntu 16.04, 18.04, and newer, Red Hat Enterprise Linux 7.6 and 8.0, SUSE, Fedora, Debian, FreeBSD, NetBSD, Amazon Corretto distribution of OpenJDK
  • Docker Desktop, Amazon ECS, Amazon EKS for containers.
  • Amazon CloudWatch, AWS Systems Manager, AWS CodeCommit, Cloud9, CodePipeline, and Amazon Inspector for tools
  • Code Suite, Jenkins

Graviton2 now has better software ecosystem than the initial AWS Arm-based A1 instances and continues to make a great use of the AWS Nitro System, an AWS platform that abstracts the underlying hardware via a thin hypervisor.

Nitro has been around for years, but AWS has rearchitected its infrastructure based on it. The upshot is that AWS can move workloads to any architecture and processor.

AWS has noted that Nitro is how they are using virtualization and optimizing the stack across hardware and software, allowing the, to modularize and build new platforms easier. ASICs take functions off core hardware and are optimized for tasks. In a nutshell, Nitro is the hypervisor created for AWS and allows the cloud provider to swap out processors and architectures easily.

Densify performs comprehensive checks leveraging our patented, rule-driven engine to compare the requirements of your workloads against the capabilities of different cloud instance types based on metadata from AWS. We make sure only suitable workloads are considered as migration candidates for M6g, and any additional considerations are highlighted and presented to our users.

Rules for evaluating workload AWS instance requirements
Examples of rules that are evaluated as part of multidimensional analysis

One of the workload analysis rules in action:

Workload analysis rule details
Workload analysis rule identifying effort impact of migration to a AWS Nitro instance

Cost Comparison of M6g vs T3, M5, C5, & R5 Instances

Let’s compare EC2 instances with 32 GB of RAM using US East (Northern Virginia) Region prices:

Instance< SizevCPUsBaseline Performance per 1 vCPUMemory (GiB)Price/Hour (Linux)Price/Hour (Windows)
M6g.2xlarge8n/a32$0.3080n/a
T3.2xlarge840%32$0.3328$0.4800
M5.2xlarge8n/a32$0.3840$0.7520
M5a.2xlarge 8n/a32$0.3440$0.7120
C5.4xlarge16n/a32$0.6800$1.4160
R5.xlarge4n/a32$0.2520$0.4360

Assuming the following:

  • Instances running @ 100% for a 1-month period
  • Baseline performance for t3.2xlarge is 3.2 vCPUs (40% * 8 vCPUs)
  • Additional charges of $0.05 (for Linux) and $0.096 (for Windows) per vCPU-hour for .6/*T3 bursting beyond credits

Here are the monthly costs for the various instance types:

Instance SizeLinux Windows
M6g.2xlarge$221.76n/a
T3.2xlarge (throttled)$239.62$325.60
T3.2xlarge (unlimited)$412.42$677.38
M5.2xlarge$276.48$541.44
M5a.2xlarge$247.68$512.64
C5.4xlarge$439.60$1019.52
R5.xlarge$181.44$313.92

Key Considerations for Running Workloads on AWS Graviton2 Arm Processors

  • The Graviton2-based instances have no Windows compatibility today
  • T3-based instances @100% will cost significantly more than the comparable M4 and R5 Instances
  • The M6g.2xlarge was the cheapest option of the compared instances that had 8 vCPUs
  • R5 is still the cheapest option to buy RAM but that probably only until R6g instances become available

Considerations for Reserved Instance & Savings Plans Purchases of Graviton2 Instances

With both Reserved Instances and Savings Plans available ,customers need to be aware of savings opportunities. With the help of Densify analytics and expert advice from our Cloud Advisors, you will be able to plan and execute comprehensive Reserved Instances and Savings Plans purchases along with Reserved Instances exchanges to increase your overall coverage and protect your existing investments.

Successfully Matching Candidate Workloads to Graviton2 Instances

Densify is the only technology with complete knowledge of available services and configurations across public cloud, private cloud, and container environments to deliver actionable and automatible optimization directives. Request a demo and see how we will deliver the most precise answer for where to place your apps and workloads in AWS.

Get Started Managing Hybrid Cloud Resources with Densify