Cutting Threads’ send latency in half

This post details optimizations made to the server-side media ingestion endpoint for Instagram Threads to reduce send latency. The primary improvement involved identifying that the write-through cache's backing store was in a faster region, causing unavoidable cross-region round trips for writes. The solution was to route requests to the faster region via a special HTTP header and routing rule, effectively batching these writes and reducing the endpoint's execution time. This resulted in an average send latency reduction of 29%.