AWS recently announced new instance types for Amazon EC2 on different processors and for different EC2 families. Most of the new instances are already available, even if only in a subset of regions; others are expected by the end of the month.
During the first week of re:invent, AWS introduced Intel Xeon M5zn instances, Graviton2-powered C6gn instances, Intel-powered D3/D3en instances, memory-optimized R5b instances and AMD-powered G4ad GPU instances.
The new M5zn Instances are a variant of the general-purpose M5 instances and use 2nd generation custom Intel Xeon Scalable (Cascade Lake) processors; they are designed for gaming, financial applications and simulation modeling applications that benefit from high per-core performances. They are built on the AWS Nitro System and are already available in most regions. Amazon claims they are the "fastest Intel Xeon Scalable CPU in the cloud".
As new instances on the Arm-based Graviton2 processors, Amazon has introduced the C6gn instances. The new instances add 100 Gbps networking and Elastic Fabric Adapter (EFA) support to the C6g family and are suggested for workloads with high networking bandwidth such as high performance computing (HPC), real-time video communications and data analytics. They are the latest iteration of the Graviton processor, started after Amazon acquired Annapurna Labs in 2015. Forrest Brazeal, senior manager at A Cloud Guru, comments:
I always used to bring up DynamoDB On-Demand Capacity as an example of how the cloud improves under the hood at no cost to customers. Lately, I’ve started to mention AWS Graviton more in that context. AWS’s chip-level innovations let them provide faster, cheaper EC2 instances as first-class citizens in the console, just a click away – no re-architecture required. This is a level of innovation totally beyond what individual customers could build on their own. You can only get this kind of benefit at cloud scale.
Among the memory intensive R5 types, the new R5b instances are designed for large relational database workloads and ERP systems. They are powered by the AWS Nitro system and offer up to 60Gbps of EBS bandwidth and 260,000 I/O operations per second (IOPS). According to AWS, they have the best network-attached storage performance among EC2 instances. They are also available on Amazon RDS for Oracle and Amazon RDS for SQL Server to support large managed commercial database applications and are certified for production SAP workloads and in-memory SAP HANA databases.
Characterized by large amounts of on-instance HDD storage, the new D3 / D3en instances are Intel-powered storage-optimized instances and the first major update of the dense storage types since 2015. They offer up to 32 vCPUs and 48 TB of storage and target customers who need massive amounts of cheaper on-instance HDD storage. They are recommended for data warehouses, data lakes, network file systems and Hadoop clusters.
The G4ad instances are the new type in the G4 family and run AMD GPUs for graphics intensive applications such as virtual workstations, game streaming and graphics rendering. Steve Roberts, developer advocate at AWS, explains when the new G4ad are a better choice than the existing G4dn instances:
G4dn instances will continue to be the best option for small-scale machine learning (ML) training and GPU-based ML inference due to included hardware optimizations like Tensor Cores. Additionally, G4dn instances are still best suited for graphics applications that need access to NVIDIA libraries such as CUDA, CuDNN, and NVENC. However, when there is no dependency on NVIDIA’s libraries, we recommend customers try the G4ad instances to benefit from the improved price and performance.
The G4ad instances are already available in a subset of regions.