As user acquisition models forecast rapid growth, a single master database will quickly become a CPU bottleneck. Today, we mapped out our horizontal scaling strategy. We are introducing read-replicas into our database cluster. All write operations (creating posts, adding likes, leaving comments) are routed to the primary Master node to ensure ACID transactional integrity. Simultaneously, the Master asynchronously replicates data to multiple Read nodes. Our API traffic is then load-balanced, directing heavy feed-fetch queries exclusively to the Read replicas. This separation of concerns allows us to scale read capacity infinitely without locking up our write threads.