Amazon ElastiCache

Amazon ElastiCache

Amazon ElastiCache is a fully managed in-memory caching service that supports two open-source in-memory caching engines: Redis (and now Valkey) and Memcached. It improves application performance by retrieving data from fast, managed, in-memory caches instead of relying entirely on slower disk-based databases.

Engines

1. Redis (and Valkey):

   • Features: Supports complex data structures (sorted sets, hashes, lists), persistence (AOF/RDB), Pub/Sub, Multi-AZ (Automatic Failover), and Read Replicas (Cluster Mode).
   • Use Case: Leaderboards, geospatial data, complex caching, session stores, chat apps.
   • Valkey: A Linux Foundation open-source evolution of Redis OSS, priced lower than Redis.
2. Memcached:

   • Features: Simple key-value store, multi-threaded (scales up well on large multicore nodes). No persistence, no replication (in the traditional sense), no Multi-AZ failover (nodes are replaced if they fail).
   • Use Case: Simple object caching, offloading DBs (HTML fragments, API calls).

Key Features

• Extreme Performance: Microsecond latency for read-heavy workloads.
• fully Managed: Automates patching, failure detection, and backups (Redis only).
• Serverless: New option that automatically scales capacity based on demand without managing nodes.
• Global Datastore (Redis): Replicate data across regions for low latency and DR.
• Data Tiering: Use local SSDs to store cooling data (Redis only) for cost savings.

Exam Tips

• "High performance / Microsecond latency": Think ElastiCache or DynamoDB Accelerator (DAX). If it's for a relational DB (RDS), it's ElastiCache. If it's for DynamoDB, it's DAX.
• Redis vs Memcached:

  • Redis: Complex data types, persistence, high availability (Multi-AZ), backup/restore, sorted sets (leaderboards).
  • Memcached: Simple Key-Value, Multi-threaded, pure caching (no persistence).
• Session State: Excellent use case for ElastiCache (share sessions across stateless web servers).
• Lazy Loading vs Write Through:

  • Lazy Loading: Load data into cache only when necessary (can cause stale data).
  • Write Through: specialized code adds data to cache whenever it writes to DB (data never stale, but write penalty).

Common Use Cases

• Database caching (RDS).
• User session storage (ShoppingCart).
• Real-time gaming leaderboards (Redis Sorted Sets).
• Geospatial queries.