Understanding AWS EC2 Instance Types and Naming Conventions

AWS offers dozens of Elastic Compute Cloud (EC2) instance types, grouped by compute, memory, storage, and other traits. Once you can read an instance name, you can tell most of what an instance offers before opening the spec sheet.

Instance Type Naming Convention

Each part of an instance name tells you something: the family, generation, processor family, additional capabilities, and size.

Take ‘c7gn.xlarge’:

• c:

Instance family (Compute optimized)

• 7:

Instance generation

• g:

Processor family (AWS Graviton processors)

• n:

Additional capabilities (Network and EBS optimized)

• xlarge:

Instance size

The sections below cover each part.

Instance Families

The first position names the instance family, which tells you the workload it targets. Common families:

• C (Compute optimized):

For compute-bound applications that need high-performance processors, such as batch processing and media transcoding.

• D (Dense storage):

For applications that need high storage density, such as large data warehouses or Hadoop clusters.

• F (FPGA):

For applications that use field-programmable gate arrays, such as hardware acceleration for specific computations.

• G (Graphics intensive):

For graphics and video processing, gaming, and 3D rendering.

• Hpc (High performance computing):

For high-performance computing in science, engineering, and finance.

• I (Storage optimized):

For workloads that need high sequential read and write access to very large data sets on local storage.

• Inf (AWS Inferentia):

For machine learning inference.

• M (General purpose):

A balance of compute, memory, and networking for a wide mix of workloads.

• Mac (macOS):

Runs macOS, for developers building for the Apple ecosystem.

• P (GPU accelerated):

For general-purpose GPU compute.

• R (Memory optimized):

For memory-bound applications that need larger memory sizes.

• T (Burstable performance):

A baseline level of CPU performance with the ability to burst higher when the workload needs it.

• Trn (AWS Trainium):

For machine learning training.

• U (High memory):

A large memory footprint per vCPU.

• VT (Video transcoding):

For video transcoding workloads.

• X (Memory intensive):

For memory-intensive applications such as high-performance databases.

Processor Families

The letter after the generation number names the processor family:

• a (AMD processors):

AMD processors.

• g (AWS Graviton processors):

ARM-based AWS Graviton processors.

• i (Intel processors):

Intel processors.

Additional Capabilities

Letters after the processor family mark extra capabilities:

• d (Instance store volumes):

Includes instance store volumes for temporary storage.

• n (Network and EBS optimized):

Higher networking and EBS performance.

• e (Extra storage or memory):

Extra storage or memory capacity.

• z (High performance):

Higher CPU performance.

• flex (Flex instance):

Flexible options for varying workloads.

Instance Sizes

The size comes after the period and ranges from ‘nano’ to very large sizes such as ‘48xlarge’, plus ‘metal’ for bare metal instances. Not every family offers every size. The size sets how much CPU, memory, and storage the instance gets.

Conclusion

Read the name left to right: family for the workload, generation for age, processor letter for the chip, extra letters for add-ons, and size for capacity. With that, you can narrow the list to a few candidates and then compare price and benchmarks.