You have learned that I/O operations are one of the primary bottlenecks in the operation of a database. Indexes help to reduce the number of I/O operations, but you can further enhance the performance of data retrieval by separating indexes from their associated tables.
We discussed how disk I/O is the slowest operation in a computer, because of the need for the physical movement of the disk head. There is another factor that contributes to slow disk performance and disk contention
The disk head is not only moving back and forth across the disk, but it is doing this in response to many requests for many types of information.
Because the disk head can read only one piece of information at a time, these multiple requests all contend for the operations of the disk head, creating contention
that forces disk requests to queue up for access. The more contention, the longer the queues, and the slower the disk access performance.
When you access data through an index, you are guaranteed to request data from the index and from the database table. If these two structures are on separate disks, you can reduce disk head contention and improve I/O performance.