Memory Optimized

Memory optimized instances are designed to offer fast performance for workloads processing large-scale data in memory. These instances are suitable for high-performance relational databases, in-memory caches, real-time processing applications for large-scale unstructured data on Hadoop and Spark clusters, High-Performance Computing (HPC), and Electronic Design Automation (EDA) applications.

Applicable type: r2a

r2a

r2a instance utilizes 3rd generation AMD EPYC 7000 series processors and memory optimization technologies to provide high performance and reliability. These instances support an all-core turbo clock speed of up to 3.6GHz and are optimized for memory-intensive workloads processing large-scale data.

Hardware specifications

Up to 3.6GHz 3rd generation AMD EPYC processor (AMD EPYC 7643)
Up to 25Gbps network bandwidth (Bare Metal Server is 50Gbps)
Instance sizes support up to 96 vCPUs and 768GiB of memory (Bare Metal Server supports 192 vCPUs and 2,048GiB of memory)
All sizes have an 8:1 Memory to vCPU ratio (excluding Bare Metal Server)
Support for AMD instruction sets (AVX, AVX2)

Use cases

High-performance databases (relational and NoSQL)
Distributed web-scale in-memory caches like Memcached and Redis
Real-time big data analytics such as Hadoop and Spark clusters

Details

Instance size	vCPU	Memory (GiB)	Network bandwidth (Gbps)
`r2a.large`	2	16	Up to 10
`r2a.xlarge`	4	32	Up to 10
`r2a.2xlarge`	8	64	Up to 10
`r2a.4xlarge`	16	128	Up to 10
`r2a.8xlarge`	32	256	Up to 10
`r2a.12xlarge`	48	384	12.5
`r2a.16xlarge`	64	512	12.5
`r2a.24xlarge`	96	768	25
`r2a.baremetal`	192	2,048	50

NUMA topology

In Non-Uniform Memory Access (NUMA) architecture, each CPU can access its own allocated memory (local memory). Using NUMA architecture allows for sharing memory across multiple processors, thus achieving high scalability.

r2a instances have the following NUMA topology according to the instance size.