AWS Caching Strategies: Optimizing Performance and Cost

This guide explores the mechanisms and services used to optimize data retrieval and application performance within the AWS ecosystem. From edge locations to in-memory databases, caching is a critical pillar of high-performing architectures.

Learning Objectives

After studying this guide, you should be able to:

• Differentiate between the various levels of caching (Edge, Database, Application).
• Evaluate the use cases for Amazon ElastiCache engines (Redis vs. Memcached).
• Design architectures that leverage Amazon CloudFront for content delivery.
• Determine when to use Read Replicas versus in-memory caching for database optimization.
• Understand the impact of Time to Live (TTL) on data consistency and performance.

Key Terms & Glossary

• TTL (Time to Live): The duration for which a cached object remains valid before it is flushed or refreshed from the source.
• Edge Location: A site that CloudFront uses to cache copies of your content closer to your users for lower latency.
• Cache Hit: When a requested data item is found in the cache, resulting in a fast response.
• Cache Miss: When requested data is NOT in the cache, requiring a trip to the origin (slower).
• Origin: The primary source of truth for data, such as an S3 bucket or an EC2 web server.
• Sharding: A method of horizontally partitioning data across multiple nodes to improve performance.

The "Big Idea"

[!IMPORTANT] The core philosophy of caching is Distance Reduction. Whether that distance is physical (geographical latency) or logical (compute cycles to query a disk), caching moves frequently accessed data into high-speed memory or geographically closer locations. This reduces the load on "heavy" back-end systems and dramatically improves the end-user experience.

Formula / Concept Box

Hierarchical Outline

1. Edge Caching (Amazon CloudFront)
   • CDN Mechanism: Distribution of content to 400+ Edge Locations.
   • Content Types: Static (images/JS) and Dynamic (API acceleration).
2. Database Caching (Amazon ElastiCache)
   • Redis: Supports complex data types, persistence, and high availability.
   • Memcached: Simple, multi-threaded, non-persistent key-value store.
3. Database-Level Optimization
   • Read Replicas: Offloading read queries from the primary DB instance.
   • DAX (DynamoDB Accelerator): In-memory cache for DynamoDB tables.
4. Application Caching
   • Local Caching: Storing data in EC2 instance memory.
   • Reverse Proxies: Using tools like Varnish on EC2 to intercept requests.

Visual Anchors

The Caching Request Lifecycle

Geographical vs. In-Memory Caching

Definition-Example Pairs

• Read Replica: An exact, read-only copy of a database instance.
  • Example: A news website experiences high traffic on articles. The primary DB handles writes (new articles), while 5 Read Replicas handle the thousands of daily read requests for existing articles.
• Lazy Loading: A caching strategy where data is only loaded into the cache when it is requested and results in a miss.
  • Example: A profile page only caches user data the first time a user logs in, rather than pre-loading every user in the system.
• Persistence: The ability of a cache to save data to disk so it survives a reboot.
  • Example: Using Redis to store shopping cart sessions so that if the cache node restarts, the users don't lose their items.

Worked Examples

Scenario 1: Scaling a Read-Intensive SQL Database

Problem: An RDS MySQL instance is hitting 90% CPU utilization due to a massive increase in "SELECT" queries from a reporting tool.

Solution Step-by-Step:

1. Analyze the Workload: Identify that the bottleneck is "Read" traffic, not "Write" traffic.
2. Deploy Read Replicas: Create up to 5 RDS Read Replicas in different Availability Zones.
3. Update Application: Change the reporting tool's connection string to point to the Reader Endpoint instead of the primary instance endpoint.
4. Result: The primary DB CPU drops to 30%, and reporting queries run faster across multiple replicas.

Scenario 2: Accelerating a Global Media Site

Problem: Users in London experience 2-second latency when downloading images stored in an S3 bucket located in the US-East-1 region.

Solution Step-by-Step:

1. Create CloudFront Distribution: Set the S3 bucket as the Origin.
2. Configure TTL: Set a TTL of 86,400 seconds (24 hours) for image files.
3. DNS Update: Point the website's image URL (e.g., images.mysite.com) to the CloudFront distribution domain.
4. Result: Subsequent users in London download the image from a London Edge Location in <100ms.

Checkpoint Questions

1. Which ElastiCache engine should you choose if you need to maintain a leaderboard that requires sorted sets?
   • Answer: Redis (supports complex data types like sorted sets).
2. How many Read Replicas can you create for a standard Amazon RDS MySQL instance?
   • Answer: 5.
3. What is the main benefit of using a CDN like CloudFront for static content?
   • Answer: Reducing latency by serving content from edge locations closer to the user.
4. True or False: Memcached supports data persistence and snapshots.
   • Answer: False (Memcached is purely in-memory and volatile).
5. What happens to a cached object when its TTL reaches zero?
   • Answer: It is considered expired/stale and will be fetched from the origin upon the next request.

Comparison: Redis vs. Memcached