One simple solution is using cache or RAID mechanisms (parallelism). But anyway there is no correct answer for this questions, it depends of your specific problem or workload.

Split the DB onto multiple physical drives or better still, you split the table on multiple drives. In that way, the speed of I/O operations is improved significantly since more heads fetch data concurrently.

This was a systems design questions that was purposefully open-ended. There's no real right answer, only more depth into the question itself.