We will revisit this topic a couple of times. As an introduction for large databases it is often true that a large portion of the time is spent waiting for the disks. There are several methods that may come to mind to improve the performance characteristics of the disk IO. Some of them are splitting tables and indexes onto separate disks, preordering the data, and using a RAID controller to stripe the data (RAID 0).

By default all tables and indexes live in the database directory. You can use tablespace and indexspace to create new tables and indexes in different directories. This will give you many of the advantages of RAID 0, with the added control of deciding what should go on each disk.