Optical clustering attempts to store logically related TEXT or BYTE data objects together on the same volume. In applications where related TEXT or BYTE data objects are often retrieved together, optical clustering minimizes time-consuming platter exchanges on the drives. This practice improves performance.
Optical clustering does not imply that TEXT or BYTE data objects that share a particular key value are stored next to each other on the optical platter. Instead, each optical cluster represents a reservation of some number of kilobytes (the clustersize) on a volume. However, the reserved kilobytes and the TEXT or BYTE data objects that are eventually stored might not be contiguous. The benefit of clustering is that you can find related TEXT or BYTE data objects on the same volume far faster than you can find them on separate volumes or platters.
Figure 2 illustrates the difference between sequential and clustered organization. In this example, the optical platters store TEXT or BYTE data objects that are scanned pictures of merchandise. The pictures belong to the cat_descr column in the stores_demo database. If you cluster the pictures, you must cluster them by reference to some other column, such as the manufacturer's name, as pictures do not have an inherent order themselves.
When stored sequentially, the optical-storage subsystem writes the pictures to the optical platter in the same order that they are inserted in the table. If you cluster them by the manufacturer code (the cluster key), the pictures for a given manufacturer are stored together. The clustered organization reduces platter exchanges on the optical drives.