AWS, the biggest hyperscaler, has been creating an amazing amount of transformational innovation year after year. This innovation hasn’t only been limited to the vast panoply of cloud services that AWS has developed over the years. This now extends into custom chips that compete with the leading x86 processor makers Intel and AMD. We see now customers beginning to deploy workloads on custom silicon developed by AWS. The Graviton line of processors developed by AWS which use 64 bit Arm core now power EC2 instances and provide customers additional choice beyond the traditionally supported Intel and AMD processors.
While Intel and AMD manufactured x86 processors dominate in the enterprise server space, ARM-based processors rule in the mobile and IoT space. The vast majority of tablets, smartphones, consumer devices, smartphones and other mobile devices run ARM chips as opposed to the x86 chips from Intel and AMD. This is due to reasons of efficiency for these small(er) form factors. As benchmarks have shown, ARM processors are not only cheaper to design and integrate but also offer higher performance characteristics. Furthermore, manufacturers have started to integrate GPUs with chips based on 32-bit ARMv7 and 64-bit ARMv8 instruction sets. These will have a range of applications in verticals such as Telco/5G, Manufacturing, Retail and Banking etc.
To rewind back in time, in 2018, AWS first introduced Graviton processors with support for EC2 A1 instances. The Graviton (1) based A1 server was the first-ever Advanced RISC Machine (ARM) based instance on AWS. The goal was to provide customers with a choice for running general-purpose workloads such as traditional n tier applications, data processing as well as scale-out workloads such as microservices applications running on (docker) containers, and really, any workloads that can run on smaller cores within a given memory footprint.
Graviton2 offers a big leap over the first generation Graviton processors. Firstly, they power a wider variety of EC2 instance types – T4g, M6g, C6g, and R6g instances, and their variants with local NVMe-based SSD storage. The goal is to provide up to 40% better price performance over comparable current-generation x86-based instances for a wide variety of workloads, including application servers, micro-services, high-performance computing, electronic design automation, gaming, open-source databases, and in-memory caches. [1]
From a metrics standpoint, the AWS Graviton2 processors also provide enhanced performance for video encoding workloads, hardware acceleration for compression workloads, and support for CPU-based machine learning inference. They deliver 7x more performance, 4x more compute cores, 5x faster memory, and 2x larger caches.[1]
From a software workload standpoint, any workload designed for Linux operating systems such as open source architectures built on Java, Python, LAMP etc are prime targets to be run on the Graviton2 based instances. AWS has published the below handy reference for customers considering Graviton2 [2].
The sweet spot for the Graviton2 based instance types will be areas such as cloud-native workloads, SaaS workloads, DevOps pipelines, IoT/ML-based use cases as well as HPC workloads. Really, any areas of the business which demand support for scale-out architectures can benefit from Graviton2.
The new Graviton2 instances boast of a comprehensive list of software supported both from AWS as well as partners –
AWS Graviton2 processors, based on the 64-bit Arm architecture, are supported by popular Linux operating systems including Amazon Linux 2, Red Hat, SUSE, and Ubuntu. Many popular applications and services from AWS and Independent Software Vendors also support AWS Graviton2-based instances, including Amazon ECS, Amazon EKS, Amazon ECR, Amazon CodeBuild, Amazon CodeCommit, Amazon CodePipeline, Amazon CodeDeploy, Amazon CloudWatch, Crowdstrike, Datadog, Docker, Drone, Dynatrace, GitLab, Jenkins, NGINX, Qualys, Rancher, Rapid7, Tenable, and TravisCI. (Source: AWS)
A key thing to mention is that all Graviton2 instances are built on the AWS designed hardware & software which is based on the Nitro system. The advantage of Nitro is that it provides not only strong multi-tenancy but also high speed local storage as well as private networking. The core value of Nitro is that it uses a thin hypervisor to abstract the underlying hardware. At the time of writing, Nitro System instances provide upto 19 Gbps EBS and upto 25 Gbps network bandwidth.
AWS’s vision for Nitro is a complete hardware/software stack (primarily the hypervisor) that enables modularity in building platforms. Nitro deserves its own set of blogposts but the three major components of Nitro as shown below are gradually beginning to show up across the EC2 instance type family – the Nitro Hypervisor, I/O Accelerator and Security Chip. The modularity of Nitro allows AWS to build composable hardware by swapping out processors and cards as dictated by workloads.
Most enterprise IT executives are under pressure to decrease operating budgets and server costs are a good portion of these budgets. The Graviton series of ARM processors can help enterprises optimize for the best price/performance ratio. Expect to see this segment of AWS hit astronomical growth in 2021 and beyond as customers begin moving workloads off of x86 chipsets to these instance types.
References
[1] AWS Graviton Processors – https://aws.amazon.com/ec2/graviton/
[2] Enable up to 40% better price-performance with AWS Graviton2 based Amazon EC2 instances – https://pages.awscloud.com/rs/112-TZM-766/images/2020_0501-CMP_Slide-Deck.pdf