Skip to content

AWS Compute Services

Amazon EC2 (Elastic Compute Cloud)

Overview

Amazon EC2 provides scalable computing capacity in the AWS cloud. It eliminates the need to invest in hardware upfront and allows you to launch virtual servers (instances) as needed.

Key Features

  • Virtual Servers: Launch instances in minutes
  • Scalable: Scale up or down based on demand
  • Flexible: Choose from various instance types
  • Secure: Integrated with VPC and security groups
  • Cost-Effective: Pay only for what you use

Instance Types

General Purpose

  • T4g, T3, T2: Burstable performance instances
  • Use cases: Web servers, small databases, development environments

  • CPU Credits: Accumulate credits when below baseline performance

  • M6i, M5, M4: Balanced compute, memory, and networking

  • Use cases: Web applications, microservices, enterprise applications
  • Characteristics: Balanced resources for diverse workloads

Compute Optimized

  • C6i, C5, C4: High-performance processors
  • Use cases: High-performance web servers, scientific modeling, batch processing
  • Characteristics: High ratio of compute to memory

Memory Optimized

  • R6i, R5, R4: Memory optimized
  • Use cases: In-memory databases, real-time analytics

  • Characteristics: High memory-to-vCPU ratio

  • X1e, X1: High memory instances

  • Use cases: In-memory databases like SAP HANA, Apache Spark

  • Characteristics: Highest memory per vCPU ratio

  • z1d: High frequency and NVMe SSD

  • Use cases: Electronic Design Automation (EDA), relational databases
  • Characteristics: High frequency processors with NVMe SSD

Storage Optimized

  • I4i, I3: NVMe SSD storage
  • Use cases: NoSQL databases, data warehousing, search engines

  • Characteristics: High random I/O performance

  • D3, D2: Dense HDD storage

  • Use cases: Distributed file systems, data processing workloads
  • Characteristics: High density HDD storage

Accelerated Computing

  • P4, P3: GPU instances for machine learning
  • Use cases: Machine learning training, high-performance computing

  • Characteristics: NVIDIA Tesla GPUs

  • G4: GPU instances for graphics workloads

  • Use cases: Game streaming, 3D visualization, video processing

  • Characteristics: NVIDIA T4 GPUs

  • F1: FPGA instances

  • Use cases: Hardware acceleration, real-time video processing
  • Characteristics: Field Programmable Gate Arrays

Pricing Models

On-Demand Instances

  • Description: Pay for compute capacity by hour or second
  • Use Cases: Short-term, irregular workloads
  • Benefits: No long-term commitments, launch/terminate anytime
  • Billing: Per-second billing (minimum 60 seconds)

Reserved Instances (RIs)

  • Description: Commit to 1 or 3-year terms for significant discounts
  • Discount: Up to 75% compared to On-Demand
  • Types:
  • Standard RIs: Up to 75% discount, can modify AZ, instance size, networking type
  • Convertible RIs: Up to 54% discount, can change instance family, OS, tenancy
  • Scheduled RIs: For predictable recurring schedules

Spot Instances

  • Description: Use spare EC2 capacity at up to 90% discount
  • Use Cases: Fault-tolerant, flexible applications
  • Interruption: Can be terminated with 2-minute notice
  • Best For: Batch jobs, data analysis, image processing

Dedicated Hosts

  • Description: Physical EC2 server dedicated for your use
  • Use Cases: Licensing requirements, compliance, security
  • Benefits: Visibility into host-level resources, BYOL (Bring Your Own License)
  • Pricing: Per-host billing

Dedicated Instances

  • Description: Instances run on hardware dedicated to single customer
  • Use Cases: Compliance requirements, licensing restrictions
  • Difference from Dedicated Hosts: No visibility into underlying hardware

Storage Options

Instance Store (Ephemeral Storage)

  • Description: Temporary block-level storage
  • Characteristics: High performance, directly attached to host
  • Limitation: Data lost when instance stops/terminates
  • Use Cases: Temporary storage, caches, scratch data

Amazon EBS (Elastic Block Store)

  • Description: Persistent block storage volumes
  • Types:
  • gp3/gp2: General Purpose SSD
  • io2/io1: Provisioned IOPS SSD
  • st1: Throughput Optimized HDD
  • sc1: Cold HDD
  • Features: Snapshots, encryption, Multi-Attach

Security Features

Security Groups

  • Description: Virtual firewalls controlling traffic
  • Default: Deny all inbound, allow all outbound
  • Rules: Based on protocol, port, source/destination
  • Stateful: Return traffic automatically allowed

Key Pairs

  • Description: Public/private key pairs for secure login
  • SSH: Used for Linux instances
  • RDP: Used for Windows instances
  • Best Practice: Use separate key pairs for different environments

IAM Roles for EC2

  • Description: Assign permissions to EC2 instances
  • Benefits: No need to store credentials on instances
  • Temporary: Automatically rotated credentials
  • Best Practice: Use roles instead of access keys

High Availability and Scaling

Auto Scaling

  • Description: Automatically adjust number of instances
  • Components:
  • Launch Template/Configuration: Instance specifications
  • Auto Scaling Group: Collection of instances
  • Scaling Policies: Rules for scaling actions
  • Benefits: Cost optimization, high availability, fault tolerance

Elastic Load Balancing

  • Application Load Balancer (ALB): Layer 7 load balancing
  • Network Load Balancer (NLB): Layer 4 load balancing
  • Classic Load Balancer (CLB): Legacy load balancer
  • Gateway Load Balancer (GWLB): Third-party virtual appliances

Placement Groups

  • Cluster: Low latency, high network performance
  • Partition: Large distributed/replicated workloads
  • Spread: Small number of critical instances

AWS Lambda

Overview

AWS Lambda is a serverless compute service that runs code without provisioning or managing servers. You pay only for the compute time consumed.

Key Features

  • Serverless: No server management required
  • Event-Driven: Triggered by events from other AWS services
  • Automatic Scaling: Scales automatically based on demand
  • Pay-per-Use: Pay only for compute time used
  • Multiple Languages: Support for various programming languages

Supported Runtimes

  • Python: 3.7, 3.8, 3.9, 3.10, 3.11
  • Node.js: 14.x, 16.x, 18.x, 20.x
  • Java: 8, 11, 17, 21
  • C#: .NET Core 3.1, .NET 6, .NET 8
  • Go: 1.x
  • Ruby: 2.7, 3.2
  • Custom Runtime: Using Runtime API

Key Concepts

Function

  • Definition: Code and configuration that processes events
  • Components: Code, runtime, handler, environment variables
  • Limits: 15-minute maximum execution time, 10GB memory

Event Sources

  • AWS Services: S3, DynamoDB, Kinesis, SNS, SQS, API Gateway
  • Direct Invocation: AWS CLI, SDKs, Lambda console
  • Schedule: CloudWatch Events/EventBridge cron expressions

Execution Environment

  • Container: Secure, isolated execution environment
  • Cold Start: First invocation or after period of inactivity
  • Warm Start: Reuse existing execution environment
  • Provisioned Concurrency: Pre-initialized execution environments

Pricing Model

  • Requests: First 1 million requests per month free, then $0.20 per 1M requests
  • Duration: Charged per 1ms, based on memory allocated
  • Free Tier: 1M free requests and 400,000 GB-seconds per month

Use Cases

  • Real-time File Processing: Process files uploaded to S3
  • Data Transformation: Transform data in streams
  • Web Applications: Serverless web backends
  • Chatbots: Process messaging platform events
  • Scheduled Tasks: Automated backup, cleanup operations

Best Practices

  • Optimize Memory: Right-size memory allocation
  • Minimize Cold Starts: Keep functions warm, use provisioned concurrency
  • Environment Variables: Store configuration outside code
  • Error Handling: Implement proper error handling and retries
  • Monitoring: Use CloudWatch for monitoring and logging

Amazon ECS (Elastic Container Service)

Overview

Amazon ECS is a fully managed container orchestration service that helps you deploy, manage, and scale containerized applications.

Key Components

Task Definition

  • Description: Blueprint for your application
  • Contents: Container images, CPU/memory requirements, networking
  • Versioning: Immutable, new versions created for changes

Service

  • Description: Runs and maintains desired number of tasks
  • Features: Load balancing, auto scaling, rolling deployments
  • Health Checks: Automatic replacement of unhealthy tasks

Cluster

  • Description: Logical grouping of compute resources
  • Types: EC2 clusters, Fargate clusters
  • Scaling: Add/remove instances based on demand

Launch Types

EC2 Launch Type

  • Description: Run containers on EC2 instances
  • Management: You manage the underlying EC2 instances
  • Use Cases: More control over infrastructure, cost optimization
  • Responsibility: Instance patching, scaling, monitoring

Fargate Launch Type

  • Description: Serverless compute for containers
  • Management: AWS manages the underlying infrastructure
  • Use Cases: Focus on application, not infrastructure
  • Benefits: No instance management, automatic scaling

Integration with AWS Services

  • Application Load Balancer: Distribute traffic across containers
  • CloudWatch: Monitoring and logging
  • IAM: Task-level permissions
  • ECR: Container image registry
  • VPC: Network isolation and security

Use Cases

  • Microservices: Break applications into smaller services
  • Batch Processing: Run batch jobs on containers
  • Machine Learning: Deploy ML models in containers
  • Web Applications: Host scalable web applications

Amazon EKS (Elastic Kubernetes Service)

Overview

Amazon EKS is a managed Kubernetes service that makes it easy to run Kubernetes on AWS without needing to install and operate your own Kubernetes control plane.

Key Features

  • Managed Control Plane: AWS manages Kubernetes masters
  • High Availability: Control plane across multiple AZs
  • Security: Integrated with IAM and VPC
  • Compatibility: Certified Kubernetes conformant
  • Automatic Updates: Managed updates and patches

Components

Control Plane

  • Management: Fully managed by AWS
  • High Availability: Runs across multiple AZs
  • Scaling: Automatic scaling based on demand
  • Updates: Managed updates and patches

Worker Nodes

  • Options: EC2 instances, Fargate, or both
  • Management: Customer managed (EC2) or AWS managed (Fargate)
  • Auto Scaling: Kubernetes Cluster Autoscaler
  • Node Groups: Managed groups of EC2 instances

Fargate for EKS

  • Serverless: Run pods without managing nodes
  • Isolation: Each pod runs in its own compute environment
  • Scaling: Automatic scaling based on pod requirements
  • Billing: Pay per pod resource consumption

Use Cases

  • Container Orchestration: Advanced container management
  • Hybrid Deployments: Consistent Kubernetes across environments
  • Microservices: Service mesh and advanced networking
  • Machine Learning: Kubernetes-native ML workflows

Integration with AWS Services

  • ALB: Application Load Balancer for Ingress
  • IAM: Service accounts for pods (IRSA)
  • VPC CNI: Native VPC networking
  • ECR: Container image registry
  • CloudWatch: Container insights and logging

AWS Batch

Overview

AWS Batch enables you to run batch computing workloads on the AWS Cloud by dynamically provisioning the optimal quantity and type of compute resources.

Key Components

Job Definition

  • Description: Template for job execution
  • Parameters: vCPUs, memory, IAM role, container image
  • Types: Single-job, multi-node parallel jobs

Job Queue

  • Description: Queue where jobs are submitted
  • Priority: Numerical priority for job execution
  • Association: Linked to compute environments

Compute Environment

  • Managed: AWS manages EC2 instances
  • Unmanaged: You manage EC2 instances
  • Fargate: Serverless compute option
  • Spot: Use Spot instances for cost optimization

Use Cases

  • Financial Risk Modeling: Monte Carlo simulations
  • Drug Discovery: Molecular modeling
  • Image/Video Processing: Media rendering
  • Scientific Computing: Genomics, weather modeling

AWS Lightsail

Overview

Amazon Lightsail is designed to be the easiest way to launch and manage a virtual private server with AWS. It includes everything you need to jumpstart your project.

Key Features

  • Simple: Easy-to-use interface
  • Predictable Pricing: Fixed monthly pricing
  • Pre-configured: Templates for common applications
  • Integrated: CDN, DNS, load balancer included
  • Scalable: Can migrate to EC2 when needed

Use Cases

  • Simple Web Applications: WordPress, LAMP stack
  • Development/Test Environments: Quick environment setup
  • Small Business Websites: Cost-effective hosting
  • Learning/Experimentation: AWS introduction

Included Features

  • Compute: Virtual private server
  • Storage: SSD-based storage
  • Networking: Data transfer allowances
  • DNS: DNS management
  • Static IP: Static IP addresses
  • Monitoring: Basic CloudWatch metrics

Key Takeaways

  1. EC2: Foundation of AWS compute, flexible virtual servers with multiple pricing options
  2. Lambda: Serverless computing, event-driven, pay-per-use model
  3. ECS: Container orchestration, choice between EC2 and Fargate launch types
  4. EKS: Managed Kubernetes service with full Kubernetes compatibility
  5. Batch: Managed batch computing for large-scale parallel workloads
  6. Lightsail: Simplified VPS solution with predictable pricing
  7. Choose Right Service: Based on requirements for control, management, and complexity
  8. Cost Optimization: Multiple pricing models and options available for different use cases