AWS Solutions Architect Exam  >  AWS Solutions Architect Notes  >  : Associate Level  >  Cheat Sheet: EC2 Instance Types

Cheat Sheet: EC2 Instance Types

1. Instance Family Overview

FamilyPrimary Use Cases
General Purpose (T, M, A)Balanced compute, memory, networking; web servers, code repositories, development environments
Compute Optimized (C)High-performance processors; batch processing, media transcoding, gaming servers, scientific modeling
Memory Optimized (R, X, z, High Memory)Large datasets in memory; databases, in-memory caches, real-time big data analytics
Storage Optimized (I, D, H)High sequential read/write to local storage; data warehousing, distributed file systems, OLTP
Accelerated Computing (P, G, F, Inf, VT)Hardware accelerators (GPUs, FPGAs); machine learning, graphics workloads, video encoding

2. General Purpose Instances

2.1 T Family (Burstable Performance)

Instance TypeKey Characteristics
T4gGraviton2 ARM processors; lowest cost; up to 40% better price-performance vs T3
T3/T3aUnlimited mode available; baseline CPU with burst credits; T3a uses AMD processors (10% cost savings)
T2Previous generation; baseline CPU performance with burst credits
  • CPU Credits: Earn credits when idle, consume when bursting above baseline
  • Unlimited Mode: Can burst beyond credits but incur charges
  • Best for workloads with variable CPU usage patterns

2.2 M Family (Balanced)

Instance TypeKey Characteristics
M7gGraviton3 processors; best price-performance for general purpose; up to 25% better performance vs M6g
M6g/M6gdGraviton2 ARM processors; up to 40% better price-performance vs M5; M6gd includes NVMe SSD
M6i/M6idIntel 3rd gen Xeon; up to 15% better compute price-performance vs M5; M6id includes NVMe SSD
M6aAMD 3rd gen EPYC; up to 35% better price-performance vs M5a
M5/M5a/M5nM5a uses AMD (10% savings); M5n offers up to 100 Gbps networking

2.3 A Family (ARM-based)

Instance TypeKey Characteristics
A1Graviton processors; scale-out workloads; cost savings for ARM-compatible applications

3. Compute Optimized Instances

3.1 C Family

Instance TypeKey Characteristics
C7g/C7gd/C7gnGraviton3; up to 25% better performance vs C6g; C7gn offers up to 200 Gbps networking
C6g/C6gd/C6gnGraviton2; up to 40% better price-performance vs C5; C6gn offers up to 100 Gbps networking
C6i/C6idIntel 3rd gen Xeon; up to 15% better compute price-performance vs C5; C6id includes NVMe SSD
C6aAMD 3rd gen EPYC; up to 15% better price-performance vs C5a
C5/C5a/C5nC5a uses AMD; C5n offers up to 100 Gbps networking for HPC workloads
  • Highest ratio of vCPU to memory among instance families
  • Optimized for compute-bound applications
  • Support Enhanced Networking for low latency and high PPS

4. Memory Optimized Instances

4.1 R Family (General Memory Intensive)

Instance TypeKey Characteristics
R7g/R7gdGraviton3; up to 30% better performance vs R6g; R7gd includes NVMe SSD
R6g/R6gdGraviton2; up to 40% better price-performance vs R5; R6gd includes NVMe SSD
R6i/R6idIntel 3rd gen Xeon; up to 15% better price-performance vs R5; R6id includes NVMe SSD
R6aAMD 3rd gen EPYC; up to 35% better price-performance vs R5a
R5/R5a/R5b/R5nR5b offers EBS-optimized by default; R5n offers up to 100 Gbps networking

4.2 X Family (Extra Large Memory)

Instance TypeKey Characteristics
X2idn/X2iedn/X2ieznUp to 2 TiB RAM per vCPU; Intel 3rd gen Xeon; X2idn/X2iedn include NVMe SSD; X2iezn optimized for single-threaded performance
X2gdGraviton2; up to 1 TiB RAM; up to 40% better price-performance vs X1 for memory-intensive workloads
X1e/X1X1e offers up to 3,904 GiB RAM; X1 offers up to 1,952 GiB RAM; SAP HANA certified

4.3 z Family (High Frequency)

Instance TypeKey Characteristics
z1dUp to 4.0 GHz sustained all-core frequency; high memory per vCPU; NVMe SSD; EDA workloads

4.4 High Memory

  • u-6tb1, u-9tb1, u-12tb1, u-18tb1, u-24tb1: 6 TiB to 24 TiB of memory
  • Purpose-built for large in-memory databases (SAP HANA)
  • Available only as bare metal instances

5. Storage Optimized Instances

5.1 I Family (I/O Intensive)

Instance TypeKey Characteristics
I4iIntel 3rd gen Xeon; up to 30 TB NVMe SSD; up to 65% better compute price-performance vs I3
I3/I3enNVMe SSD storage; I3en offers up to 60 TB storage and 100 Gbps networking
  • Optimized for high random I/O performance
  • NoSQL databases, transactional databases, data warehousing

5.2 D Family (Dense Storage)

Instance TypeKey Characteristics
D3/D3enHDD-based instance storage; up to 336 TB; lowest cost per TB; MapReduce, HDFS, distributed file systems
D2Previous generation; HDD-based; up to 48 TB

5.3 H Family (HDD Optimized)

Instance TypeKey Characteristics
H1Up to 16 TB HDD storage; high disk throughput; MapReduce, HDFS

6. Accelerated Computing Instances

6.1 P Family (GPU Compute)

Instance TypeKey Characteristics
P5NVIDIA H100 GPUs; up to 8 GPUs; 640 GB GPU memory; highest ML training performance
P4d/P4deNVIDIA A100 GPUs; up to 8 GPUs; 400 Gbps networking; P4de offers 1.1 TB system memory
P3/P3dnNVIDIA V100 GPUs; up to 8 GPUs; P3dn offers 100 Gbps networking
P2NVIDIA K80 GPUs; previous generation
  • Optimized for machine learning training, HPC, deep learning
  • Support NVIDIA CUDA, cuDNN libraries

6.2 G Family (Graphics Intensive)

Instance TypeKey Characteristics
G5gGraviton2; ARM Mali GPUs; Android game streaming, graphics rendering
G5NVIDIA A10G GPUs; up to 8 GPUs; graphics workloads, ML inference, video encoding
G4dnNVIDIA T4 GPUs; up to 8 GPUs; ML inference, graphics, video transcoding
G4adAMD Radeon Pro V520 GPUs; up to 4 GPUs; graphics workloads
G3NVIDIA M60 GPUs; previous generation

6.3 F Family (FPGA)

Instance TypeKey Characteristics
F1Xilinx UltraScale+ FPGAs; up to 8 FPGAs; custom hardware acceleration, genomics, financial analytics

6.4 Inf Family (Inferentia)

Instance TypeKey Characteristics
Inf2AWS Inferentia2 chips; up to 12 accelerators; up to 4x higher throughput vs Inf1; lowest cost for ML inference
Inf1AWS Inferentia chips; up to 16 accelerators; ML inference workloads

6.5 VT Family (Video Transcoding)

Instance TypeKey Characteristics
VT1Xilinx U30 media accelerators; up to 8 accelerators; real-time video transcoding

6.6 DL Family (Deep Learning)

Instance TypeKey Characteristics
DL1Gaudi accelerators by Habana Labs; up to 8 accelerators; deep learning training

6.7 Trn Family (Trainium)

Instance TypeKey Characteristics
Trn1/Trn1nAWS Trainium chips; up to 16 accelerators; deep learning training; up to 40% cost savings vs P4d; Trn1n offers 1600 Gbps networking

7. Instance Naming Convention

ComponentMeaning
Instance Family (e.g., m, c, r)Indicates workload type (general, compute, memory, etc.)
Generation (e.g., 5, 6, 7)Higher number = newer generation
Additional Capabilitiesa=AMD, g=Graviton, n=network optimized, d=NVMe, e=extra capacity, z=high frequency, b=block storage, dn=dense storage with network
Size (e.g., nano, micro, small, medium, large, xlarge, 2xlarge)Resource scale within family
  • Example: m6i.2xlarge = M family, 6th gen, Intel, 2xlarge size
  • Example: c7g.xlarge = C family, 7th gen, Graviton, xlarge size
  • Example: r5d.4xlarge = R family, 5th gen, NVMe SSD, 4xlarge size

8. Key Selection Criteria

8.1 By Workload Type

WorkloadRecommended Family
Web servers, app serversT (variable traffic), M (steady traffic)
Batch processing, HPCC (compute optimized)
In-memory databases, cachesR, X (memory optimized)
NoSQL databases, data warehousingI (storage optimized with NVMe)
Big data, Hadoop, distributed file systemsD (dense HDD storage)
ML training (large models)P, Trn (GPU/accelerators)
ML inferenceInf, G4dn (purpose-built inference)
Graphics, video encodingG (graphics optimized)

8.2 Processor Comparison

Processor TypeKey Points
Graviton (ARM)Up to 40% better price-performance; best for scale-out workloads; requires ARM-compatible code
Intel XeonBroad software compatibility; higher single-threaded performance options
AMD EPYC10-35% cost savings vs Intel equivalents; excellent multi-threaded performance

8.3 Network Performance Tiers

  • Low to Moderate: Up to 10 Gbps (T, smaller M/C/R sizes)
  • Up to 25 Gbps: Standard enhanced networking
  • Up to 100 Gbps: High-throughput instances (C5n, M5n, R5n, I3en, P3dn)
  • Up to 200 Gbps: Latest generation (C7gn)
  • Up to 400 Gbps: P4d instances
  • Up to 1600 Gbps: Trn1n instances

8.4 Instance Storage Options

Storage TypeInstances
EBS-onlyMost instance types by default (M6i, C6i, R6i)
NVMe SSDInstances with 'd' suffix (M6gd, C6id, I4i, I3); high IOPS and throughput
HDDD3, D2, H1; high sequential throughput, lower cost per TB
  • Instance store data is ephemeral (lost on stop/termination)
  • NVMe provides lower latency than EBS

9. Purchasing Options Impact

OptionKey Characteristics
On-DemandNo commitment; pay per second (Linux) or hour (Windows); highest cost
Reserved Instances1 or 3 year commitment; up to 72% savings; standard or convertible
Savings PlansFlexible commitment to compute spend; up to 72% savings; applies across instance families
Spot InstancesUp to 90% savings; can be interrupted; fault-tolerant workloads only
Dedicated HostsPhysical server dedicated to your use; compliance, licensing requirements
Dedicated InstancesInstances run on hardware dedicated to single customer; no other customers share hardware

10. Special Considerations

10.1 Bare Metal Instances

  • Direct access to physical server; no virtualization overhead
  • Available in M5, M6i, C5, C6i, R5, R6i, I3, I4i families and High Memory
  • Use cases: workloads requiring access to hardware features, non-virtualized environments, licensing restrictions

10.2 Enhanced Networking

  • Elastic Network Adapter (ENA): Up to 100 Gbps; most current generation instances
  • Intel 82599 VF: Up to 10 Gbps; older instance types
  • Elastic Fabric Adapter (EFA): HPC and ML workloads; OS-bypass for low latency; P4d, P5, C6gn, C7gn

10.3 Placement Groups

  • Cluster: Low latency, high throughput within single AZ; HPC workloads
  • Partition: Spreads instances across logical partitions; distributed databases like Hadoop, Cassandra
  • Spread: Each instance on distinct hardware; small number of critical instances

10.4 Nitro System

  • Underlying platform for modern EC2 instances (5th gen and later)
  • Dedicated hardware for storage, networking, security
  • Benefits: better performance, enhanced security, faster innovation, lower cost
  • Enables bare metal instances and higher network speeds

10.5 Instance Size Selection

  • Start small, scale vertically based on CloudWatch metrics
  • Monitor CPU, memory, network, disk I/O utilization
  • Each size doubling: 2x vCPU, 2x memory, proportional network increase
  • Some features only available at certain sizes (e.g., enhanced networking at large and above)

11. Quick Reference: Top Instance Types

11.1 Most Common for Web Applications

  • T3/T3a: Development, test environments, low-traffic web servers
  • M6i/M6a: Production web applications, microservices
  • C6i/C6a: High-traffic web servers, API backends

11.2 Most Common for Databases

  • R6i/R6g: General purpose databases (RDS alternative)
  • X2gd/X2idn: Large in-memory databases, SAP HANA
  • I4i/I3en: High-performance NoSQL (Cassandra, MongoDB)

11.3 Most Common for Big Data

  • D3/D3en: HDFS, MapReduce data nodes
  • I4i/I3en: High-performance analytics
  • M6i/C6i: Master nodes, compute-heavy analytics

11.4 Cost Optimization Picks

  • T4g: Lowest cost general purpose (ARM)
  • M6g/C6g/R6g: Best price-performance with Graviton2
  • M6a/C6a/R6a: AMD options for 10-35% savings over Intel
  • Spot + T3/M6i/C6i: Up to 90% savings for fault-tolerant workloads
The document Cheat Sheet: EC2 Instance Types is a part of the AWS Solutions Architect Course AWS Solutions Architect: Associate Level.
All you need of AWS Solutions Architect at this link: AWS Solutions Architect
Explore Courses for AWS Solutions Architect exam
Get EduRev Notes directly in your Google search
Related Searches
Important questions, Summary, practice quizzes, Previous Year Questions with Solutions, study material, past year papers, Cheat Sheet: EC2 Instance Types, Cheat Sheet: EC2 Instance Types, Semester Notes, Sample Paper, Viva Questions, Extra Questions, mock tests for examination, shortcuts and tricks, ppt, Exam, pdf , video lectures, MCQs, Cheat Sheet: EC2 Instance Types, Objective type Questions, Free;