- based on the principle of Locality of Reference
- store copies of frequently accessed or computationally expensive data in a cache to improve performance
- can be used at all levels of architecture - frontend & backend
Global and distributed caches
When caching at a node level in Distributed Systems, if the load balancer (Load Balancing) randomly distributes requests, it leads to increased cache misses. The solution in such scenarios is to use global or Distributed Caches.
Cache Invalidation
Process of removing or updating stale cached data in a cache system to ensure that the data remains coherent with the source of truth.
Write-Around Cache
- write to database, bypassing cache
- leads to higher cache miss → higher latency for recently updated data
- cache invalidation
- time-based expiration: invalidate cache after a time
- event-based invalidation: throw event after database write
Write-Through Cache
- write both in cache and database for every write operation
- no issues in case of any system interruption
- higher latency for write operations
Write-Back Cache
- write to cache alone and send back confirmation to client
- write to database after some set intervals
- low latency for write-intensive apps
- risk of data loss - since the data is only in the cache for some time, cache is the only source of truth