Ref: https://learn.cantrill.io/courses/1820301/lectures/41301316 and https://learn.cantrill.io/courses/1820301/lectures/41301317

• Summary
• Contents:

Selecting an EC2 Instance Type

• Select a specific type = select family, generation, size & extra capabilities
  • Granular control over resource configuration
    • Pick correct type for a specific workload → well-performing system
    • Pick wrong type → bad customer experience
  • No fits-all solution & often several different types might fit same requirements
    • ❗but it's important not to select the wrong type!
• Selection of instance type influences:
  • Amount of raw resources
    • CPU, memory, local storage type, local storage capacity
  • Resource ratios
    • e.g a compute instance gives more CPU & less memory for the same cost
  • Storage & Data NW BW
    • e.g. EBS volume might need certain BW (if instance does not provide enough → instance is the limiting factor)
  • System Architecture
    • e.g. ARM architecture, x86 architecture
  • System Vendor
    • e.g. Intel CPUs, AMD CPUs
  • Additional features & capabilities
    • e.g. special types of CPUs → GPUs, FPGAs…
• 💡 Instances are very customizable compute HW

EC2 Instance Categories

💡 Categories group similar instance types together based on their purpose

1. General Purpose
   • Diverse workloads
   • Equal resource ratio
   • Default
     • Start off with general purpose instances, choose others when specific requirements come
   • Common general purpose types:
     • A1, M6g → ARM processor, efficient (small intances with low costs)
     • T3, T3a → Burst Pool (low CPU usage with occasional peaks)
     • M5, M5a, M5n → Steady state workload
2. Compute Optimized
   • Access to highest-performing CPUs
   • Ratio: +CPU, -memory
   • Scenarios: media processing, High Performance Computing (HPC), scientific modeling, gaming, Machine Learning (ML)…
3. Memory Optimized
   • Process large in-memory datasets
   • Ratio: -CPU, +memory
   • Scenarios: in-memory caching (e.g. Redis, ElastiCache), some DB workloads…
4. Accelerated Computing
   • Have extra features & capabilities
   • Scenarios: high scale parallel processing or modeling (HW GPU), custom programmable HW like Field Programmable Gate Arrays (FPGAs)…
5. Storage Optimized
   • Large amounts of super-fast local storage
   • Scenarios: high sequential transfer rates & massive amounts of IO operations per second (sequential & random IO), scale-out transactional DBs, Data Warehousing (DWH) - e.g. Amazon Redshift, Elasticsearch, data analytics workloads…

💡 Mnemonic: Great Pirates Conquer Oceans, Master Onboard Acrobatics, Collect Stolen Opulence

Decoding EC2 Instance Types

Diagram: https://github.com/acantril/aws-sa-associate-saac03/blob/main/0900-ELASTIC_COMPUTE_CLOUD(EC2)/00_LEARNINGAIDS/EC2InstanceTypes.png