Sequential Access Costs

Disk costs are lowest when the database server reads the rows of a table in physical order. When the first row on a page is requested, the disk page is read into a buffer page. Once the page is read in, it need not be read again; requests for subsequent rows on that page are filled from the buffer until all the rows on that page are processed. When one page is exhausted, the page for the next set of rows must be read in. To make sure that the next page is ready in memory, use the read-ahead configuration parameters described in RA_PAGES and RA_THRESHOLD.

When you use unbuffered devices for dbspaces, and the table is organized properly, the disk pages of consecutive rows are placed in consecutive locations on the disk. This arrangement allows the access arm to move very little when it reads sequentially. In addition, latency costs are usually lower when pages are read sequentially.