ElasticCache Strategy

Caching strategies for ElasticCache

Caching design pattern

• Lazy Loading / Cache-aside / Lazy population (all the same thing)
  • Only write to cache if cache does not contain the information
  • 
  • Pros: Only requested data is cached.
  • Cons: Cache miss resulted into 3 round trips so extra latency. Data can be stale as well in the database if cache miss keeps happening
• Write through
  • Add or update the cache when database is updated
  • 
  • Pros:
    • Data is never stale, reads are quick
    • Write penalty is only 2 calls comparing to lazy loading
  • Cons:
    • Data is missing until you update or add new data into database
      • We can also implement lazyloading as well to counter this
    • Cache churn: Add a lot but how much of the cache is actually needed
• Cache Evictions and Time-to-live (TTL)
  • Remove the data out of your cache to save space
  • Item is evicted because memory is full and it's not recently used (LRU)
  • If too many evictions happen due to memory then increase cache size

Note

• Lazy-loading works really well and easy to implement in all situation. Especially on the read side
• Write-through is usually combined with Lazy Loading to maximise the benefit
• TTL is not bad but becareful when you use write through (because you keep writing to the cache). You need to tune to a good parameter
• Only cache the data that makes sense (user profiles, blogs, etc ...)