AWS has launched Graviton4-powered memory-optimized Amazon EC2 X8g instances designed for high-memory workloads. Available in ten virtual sizes and two bare metal configurations, these instances offer up to 3 TiB of DDR5 memory and 192 vCPUs, with significant performance enhancements over previous models.
The X8g instances are built upon the 5th generation of AWS Nitro System and Graviton4 processors. These processors include extra security features such as Branch Target Identification (BTI), which safeguards against low-level attacks that aim to disrupt control flow at the instruction level. In addition, the instances support ENA, ENA Express, and EFA Enhanced Networking.
Chief Evangelist for AWS Jeff Barr writes:
When compared to the previous generation (X2gd) instances, the X8g instances offer 3x more memory, 3x more vCPUs, more than twice as much EBS bandwidth (40 Gbps vs 19 Gbps), and twice as much network bandwidth (50 Gbps vs 25 Gbps).
And:
The Graviton4 processors inside the X8g instances have twice as much L2 cache per core as the Graviton2 processors in the X2gd instances (2 MiB vs 1 MiB), 160% higher memory bandwidth, and can deliver up to 60% better compute performance.
EC2 X8g Instance Sizes (Source: AWS News blog post)
According to the company, the EC2 X8g instances have the best price performance among EC2 X-series instances. They are ideal for memory-intensive workloads such as electronic design automation (EDA) workloads, in-memory databases (Redis, Memcached), relational databases (MySQL, PostgreSQL), real-time big data analytics, real-time caching servers, and memory-intensive containerized applications.
In a Reddit tread on storage versus memory optimized EC2 instance, a respondent explained:
With EC2, consider three instance types: C, M, and R.
C is compute optimized. Every vCPU gets 2 GB of RAM, a 1:2 ratio.
M is...in the middle. 1:4 ratio.
R is memory/RAM optimized; for every 1vCPU, you get 8 GB of RAM.
This lets you pick the right size of instance for your workload without "wasting" anything. If you have a 32-bit application that can only use 4 GB of RAM, use a C-type instance. Need tons of memory for caching, but not that much compute (e.g., a database)? R-type. Not sure? M-type.
With the new instances, AWS keeps pushing the boundaries to stay ahead of its competitors, Azure and Google. Regarding memory optimized virtual machines, Microsoft offers the M-family of VM sizes, which are Azure's ultra memory-optimized VM instances, while Google also provides a memory optimized series of VMs.
Lastly, the X8g instances are available in the US East (N. Virginia), US West (Oregon), and Europe (Frankfurt) AWS regions in On-Demand, Spot, Reserved Instance, Savings Plan, Dedicated Instance, and Dedicated Host forms.