Elasticache

Introduction

• Fully managed Redis or Memcached instances
• Caches are in-memory databases with high-performance and low latency
• Helps to reduce load from databases
• Helps to make your application stateless
• Requires that your application be architected with a cache in-mind

Redis

• Supports multi-AZ with auto-failover
• Supports read-replicas to scale out reads
• Backup and restore features
• Supports sets and sorted sets

Memcached

• Multi-node for partitioning of data (Sharding)
• None of the features that Redis supports

Caching Implementation Considerations

• Is it safe to cache the data?
  • Data may be out of date (eventually consistent)
• Is caching effective for that data?
  • Patterns: data changing slowly, few keys are frequently updated
  • Anti patterns: data changing rapidly, all large key space frequently needed
• Is data structured for caching?
  • example: key value caching or caching of aggregations result

Caching design patterns

Lazy Loading / Cache-Aside / Lazy Population

Write-through

Cache Evictions and TTL

• Cache eviction can occur in 3 ways:
  • You delete the item in the cache
  • Item is evicted because the memory is full and its not in use (LRU)
  • The TTL (time to live) has expired
    • TTL can range from a few seconds to days