Skip to Content
Core InfraNode ManagementKarpenter Optimization (AWS + Azure)Instance Types

Instance Types

Overview

Cloud providers offer diverse ranges of instance types optimized for different use cases. Understanding these options is crucial for deploying cost-effective Kubernetes clusters that meet your workload requirements.

AWS EC2 Instance Types

Instance Categories

EC2 instances are organized into logical groupings based on CPU-to-memory ratios:

  • General Purpose (m category): Balanced CPU-to-memory ratio suitable for most workloads
  • Compute Optimized (c category): High CPU-to-memory ratio ideal for compute-intensive applications
  • Memory Optimized (r category): High memory-to-CPU ratio perfect for memory-intensive workloads

Instance Families

Instance families are identified by a naming convention:

  • First letter: Indicates the category (m for general purpose, c for compute, etc.)
  • Number: Indicates the generation (higher numbers represent newer generations)
  • Suffix letters: Indicate special features:
    • a: AMD processors
    • g: AWS Graviton processors (ARM-based)
    • i: Intel processors
    • d: NVMe storage
    • n: Enhanced networking capabilities

For example, m6g represents a 6th generation general purpose instance with AWS Graviton processors.

Instance Sizes

Within each family, instances are available in multiple sizes that scale proportionally:

  • Small-scale (nano, micro, small): Entry-level instances with limited resources
  • Mid-tier (medium, large, xlarge): Balanced instances for common workloads
  • Large-scale (2xlarge to 24xlarge): Powerful instances for resource-intensive applications
  • Bare metal (metal): Direct hardware access without virtualization overhead

Azure VM Instance Types

Instance Categories

Azure VMs are organized into logical groupings optimized for specific purposes:

  • General Purpose (D, B, A families): Balanced CPU-to-memory ratio for testing, development, and small to medium databases
  • Compute Optimized (F family): High CPU-to-memory ratio for compute-intensive workloads
  • Memory Optimized (M, E families): High memory-to-CPU ratio for memory-intensive workloads
  • Storage Optimized (L family): High disk throughput and I/O for storage-intensive workloads
  • GPU Accelerated (N family): Equipped with GPUs for graphics-intensive applications and deep learning
  • High Performance Compute (H family): Optimized for HPC workloads with features like RDMA networking

Instance Families

Azure VM families are identified by a naming convention:

  • Family letter: Indicates the category (D for general purpose, F for compute, M for memory, etc.)
  • Subfamily letter: Indicates specific features (s for premium storage, i for isolated instances, etc.)
  • Version number: Indicates the generation (v2, v3, v4, v5, etc.)
  • CPU indicator:
    • a: AMD processors
    • p: ARM-based processors (Microsoft Cobalt or Ampere Altra)
    • No letter: Intel x86-64 processors

For example, D8s_v5 represents a general purpose VM with 8 vCPUs, premium storage support, and Intel processors.

Instance Sizes

Azure VM sizes scale proportionally within each family:

  • Small-scale: B1s, B2s, D2s_v3
  • Mid-tier: D4s_v3, D8s_v3, F4s_v2
  • Large-scale: D16s_v3, D32s_v3, M32s_v2
  • Specialized: N-series for GPU, H-series for HPC

Karpenter Support

Manage Node PoolsMigration Examples