Instance naming conventions

更新时间:
复制 MD 格式

Elastic GPU Service (EGS) instance names encode the product, instance family, and vCPU count directly into the name. Once you understand the pattern, you can read any instance name and immediately know its GPU type, hardware architecture, and size—without consulting a spec sheet.

This topic covers naming conventions for heterogeneous computing instance families, ECS Bare Metal instance families, and Super Computing Cluster (SCC) instance families. Other instance families follow separate conventions not described here.

Name structure

Every instance type name follows this format:

ecs.<instance-family>.<nx>large
SegmentDescriptionExample
ecsProduct code for Elastic Compute Service (ECS)ecs
<instance-family>Family name body + family name suffixgn6v, ebmgn7e
<nx>largevCPU count — higher n means more vCPUs24xlarge = 96 vCPUs

vCPU count reference:

SuffixvCPUs
xlarge4
2xlarge8
3xlarge12

For larger sizes, multiply n by 4. For example, 24xlarge = 24 x 4 = 96 vCPUs.

Instance family names

An instance family name consists of two parts: a name body and a name suffix. Together they identify the GPU type, hardware architecture generation, and product line.

Name body

The name body identifies the acceleration type and product line.

Name bodyInstance type
gnNVIDIA GPU-accelerated compute-optimized
vgnNVIDIA GRID vGPU-accelerated dedicated
sgnNVIDIA GRID vGPU-accelerated shared
giIntel GPU-accelerated compute-optimized
fFPGA-accelerated compute-optimized
ebmc, ebmg, ebmr, ebmgn, ebmhfECS Bare Metal instance
sccc, sccg, scch, sccgn, scchfSuper Computing Cluster (SCC)

Name suffix

The name suffix identifies the hardware architecture generation (number) and GPU model or feature (letter).

Generation numbers:

NumberArchitecture
6Volta or Turing
7Ampere

Feature letters:

LetterMeaning
vNVIDIA V100 GPUs with 16 GB of GPU memory each
eGeneration 6 (6e): second-generation NVIDIA V100 GPUs with 32 GB of GPU memory each (short for "extend"). Generation 7 (7e): Ampere architecture (see Example 2 for details).
iNVIDIA T4 GPUs (generation 6) or NVIDIA A10 GPUs with 24 GB of GPU memory each (generation 7, short for "inference")
sSixth-generation SG-1 (generation 6) or seventh-generation NVIDIA A30 GPUs (generation 7)

For example, the suffix 6v means Volta or Turing architecture with NVIDIA V100 GPUs (16 GB). The suffix 7i means Ampere architecture with NVIDIA A10 GPUs optimized for inference.

Decode an instance name

The following examples show how to read each segment of an instance name.

Example 1: `ecs.sccgn6.24xlarge`

SegmentValueMeaning
ProductecsElastic Compute Service
Family bodysccgnGPU-accelerated compute-optimized SCC instance
Family suffix6Volta or Turing architecture
Size24xlarge24 x 4 = 96 vCPUs

Example 2: `ecs.ebmgn7e.32xlarge`

SegmentValueMeaning
ProductecsElastic Compute Service
Family bodyebmgnGPU-accelerated compute-optimized ECS Bare Metal instance
Family suffix7eAmpere architecture;
Size32xlarge32 x 4 = 128 vCPUs

Example 3: `ecs.vgn6i-m4.xlarge`

SegmentValueMeaning
ProductecsElastic Compute Service
Family bodyvgnNVIDIA GRID vGPU-accelerated dedicated instance
Family suffix6iVolta or Turing architecture; NVIDIA T4 GPU
DriverUses the NVIDIA GRID driver
Sizexlarge4 vCPUs

What's next