The partition index file has an external name in the format map–file–name_ndx.dat. It is a global index used to look up the partition in which an object is located, using its object identifier (oid). Only the instance identifier and not the class identifier is stored, since a partitioned file contains instances of a single class and its subobjects.
Locating and centralizing the indexes in a separate file provides an efficient mechanism to locate an object in a partition using its object identifier (oid), and to determine whether an object resides in an offline partition.
When a class and database file is partitioned, subobjects are located in the same partition as their parent. Subobjects are exclusive collections, Unstructured Data Resources (UDRs), binary large objects (blobs), and string large objects (slobs). To determine the partition in which a subobject is located, the instance identifier of the parent is required. The parent instance identifier is directly available in the oid for exclusive collections, blobs, and slobs and from the header oid attribute for collection blocks.