Caching strategy
The ECS architecture implements a multi-layer caching strategy to optimize performance and reduce latency.
In-memory cache
-
Each ECS task maintains local caches for transformation policies and origin mappings
-
Cache is populated during task startup from DynamoDB
-
Fast lookup times for frequently accessed configurations
Cache initialization
-
New ECS tasks query DynamoDB during startup
-
All transformation policies are loaded into memory
-
Origin mappings are stored in optimized data structures
Cache invalidation
-
Rolling deployment strategy updates all ECS tasks with new configurations
-
Eventual consistency model balances performance with data freshness
-
Cache refresh typically takes up to 5 minutes
Cache consistency
-
New ECS tasks always load the latest configuration data
-
Existing tasks maintain cached data until manual refresh or natural replacement
-
Auto-scaling events automatically provision tasks with current data
-
Stale cache scenarios are resolved through the rolling update process
