AWS Scaling Strategies

Scaling on AWS

• The presentation focuses on scaling up to 10 million users on AWS.
• The presenter, Giulio Lannazzo, is a Solution Architect at Amazon Web Services.
• The presentation covers various AWS services, infrastructure, and considerations for scaling.

How to Scale?

• Managing users, maintaining performance (including multiple geographies), detecting and responding to incidents, maintaining business continuity, and managing security and compliance are critical for scaling.
• Developing and testing, managing change, tracking and managing costs, optimizing for cost, and minimizing the carbon footprint are also important.

AWS Global Infrastructure

• AWS operates in 34 Regions.
• There are 108 Availability Zones (AZs) across these regions.
• A private backbone network connects these regions and AZs.
• AWS has over 600 points of presence.

AWS Regions

• Regions establish data locations.
• Protection against legal risks (such as GDPR) is ensured through regions.
• Disaster recovery capabilities are also supported by regions.

Availability Zones (AZs)

• Each region comprises at least three AZs.
• Each AZ houses a cluster of data centers.
• AZs provide resilience against technical failures.

Points of Presence

• Points of presence (POPs) are used for content delivery services like Amazon CloudFront.
• POPs are not directly accessible by users.

AWS Local Zones

• AWS Local Zones extend AWS's regional infrastructure to key locations. These are in major cities and industrial centers.

AWS Wavelength

• AWS Wavelength brings AWS cloud infrastructure closer to end-users and allows for low-latency access (by using 5G).
• It is tailored for applications like real-time data analysis and processing.

AWS Outposts

• Extending AWS infrastructure to your on-premises data center is enabled by AWS Outposts.
• This allows for ultra low-latency application use.
• Local data processing and storage are supported.

Services Summary

• Over 200+ fully featured AWS services are mentioned in this presentation.
• These services cover analytics, business applications, blockchain, security, identity, compliance, databases, compute, storage, media services, hybrid architecture, Internet of Things, machine learning, and more.

AWS Considerations

• Many important decisions are noted as reversible, allowing for adjustment.
• A cycle of building, measuring, and learning is highlighted.

User Capacity Considerations

• A single instance deployment shows shortcomings in terms of redundancy and scalability for single-instance deployments.
• More comprehensive solutions are necessary when multiple users are accessing the service.
• Managed services (RDS, DynamoDB) are noted as alternatives to self-managed solutions.

Scaling Considerations

• The presentation acknowledges a shift toward horizontal scaling (adding more computing resources) over vertical scaling (using more powerful single instances).
• Load balancing is discussed as a way to spread user requests across multiple instances for optimal performance.
• Application Load Balancers and Network Load Balancers are suggested.

Database Choices

• Self-managed (Amazon EC2) and fully managed (RDS, DynamoDB, Aurora, Timestream) database options are presented.
• Database performance considerations for scaling are referenced. A note on choosing NoSQL versus Relational databases.

Typical Read/Write Ratio

• A read-heavy user traffic pattern (70% reads, 30% writes) implies the benefit of read replicas for database performance.

Amazon S3

• Amazon S3 provides durable, high-capacity object storage.
• S3 is a great option for static assets, scalable to very high volumes, and able to handle object sizes up to 5TB.

Amazon CloudFront

• For high-volume content delivery and fast content delivery, Amazon CloudFront caches content close to users. It streamlines content delivery for everything from static assets and video to optimizing time to live (TTL).

Caching Considerations

• Static web content like images, videos, CSS files, and JS files are good candidates for caching.

Amazon ElastiCache

• AWS ElastiCache is identified as for caching Memcached and Redis data sets, scaling easily and providing millisecond-level speeds.

Amazon DynamoDB

• A good solution for NoSQL databases, DynamoDB offers high performance, distribution, fault tolerance, and global (multi-region) table features.

Service Oriented Architecture (SOA)

• Splitting components and scaling separately are highlighted as advantages of SOA.
• The service-oriented architecture approach. Service contracts define how services interact.
• Choosing the right runtimes (e.g., Lambda, ECS, EKS) based on application characteristics is crucial for efficiency.

Serverless Web Application Summary

• A sample serverless web application architecture is presented.

Loose Coupling Considerations

• Loose coupling (e.g., using queues like SQS and topics like SNS) allows greater scalability and resilience.
• Event-driven architectures are viable (using services like Lambda) and are emphasized.

Microservice Architecture

• A microservices architecture is highlighted as a suitable structure for larger applications, especially beyond 1,000,000 users.

AWS X-Ray

• AWS X-Ray tools help diagnose performance issues for microservice environments.

> 1 Million Users Planning Considerations

• Fine-tuning the solution, considering a multi-region deployment, and handling potential database performance issues are pointed out. Also custom tooling if needed.

Additional considerations

• Purpose-built databases (graph, ledger, geo-tracking) are suitable for specific application needs.
• Potential database issues and potential solutions are outlined.