Describe your approach to scaling a database cluster to handle millions of read/write operations per second without causing downtime.
