Some fundamental server-side latency optimizations before introducing caching :
- Reduce the payload size (avoid sending unnecessary extra fields/data over the network, fetch only the required fields instead of using SELECT *)
- Move non-critical and long running tasks out of the request flow (like say a image processing, or another external network call, analytics etc.,, handle them asynchronously)
- Use pagination when returning huge amounts of data.
- Use efficient formats wherever possible (serialization/deserialization speed matters, so formats like protobufs can be significantly faster than json)
- Use HTTP/2 or HTTP/3 where possible (they support multiplexing, and allow multiple requests over a single connection and reducing network overhead)
- Optimize slow database queries by proper database indexing, identifying and resolving N+1 queries, using connection pooling (keeps reusable TCP/DB connections warm, avoids creating new connections for every request, and improves latency)