How to scale a service with linear load growth? What to do when horizontal scaling hits database limits?

To scale a service with linear load growth, the following approaches are usually used:

1. Horizontal scaling (sharding, replication) — adding new service instances behind a load balancer.
2. Caching — using Redis, Memcached to reduce database load.
3. Query and index optimization — reducing database response time.

If horizontal scaling hits database limits (e.g., due to connection or performance restrictions), you can:

• Shard the database — distribute data across multiple databases by key (e.g., by users).
• Use CQRS (Command Query Responsibility Segregation) — separate read and write operations into different services and databases.
• Implement asynchronous processing — use message queues to offload the database.
• Switch to a more scalable DBMS — for example, NoSQL solutions or NewSQL, which scale better horizontally.

Example: if the service cannot handle load on a single database, data can be partitioned by regions, and requests directed to the corresponding shards, thereby reducing load on each individual database.