TREX differentiates between the following index types:
A physical index is the smallest index available to all the TREX standard functions. The TREX standard functions include the full-text search, attribute search, and text-mining functions.
From the perspective of the application that creates an index, the physical index is not further structured. A physical index is subdivided within TREX as follows:
For every language there is a specific language index.
The TREX engines have different requirements regarding the data structure of an index. As a result, each language index consists in turn of a full-text index, an attribute index, and a text-mining index.
The graphic below shows the internal structure of a physical index:
A logical index is an index that consists of several physical indexes and metadata. The metadata contains information about which physical indexes belong to which logical indexes. The components of the index (the metadata and physical indexes) are administered by different index servers.
A logical index is generated by TREX if a physical index can no longer be managed by a single index server due to its size.
The following figure shows a logical index and its breakdown into several index servers:
A compound index is an index that consists of several table indexes and metadata. It allows you to search for structured data that is distributed over several table indexes.
As the name 'table index' suggests, the contents of the index originate from a database table. Each table index is either a physical or logical index (depending on its size).
The metadata contains information about which table indexes belong to which compound index and how to link the table indexes together.
The graphic below shows a compound index:
A table index can belong to several compound indexes. In this respect, a table index differs from a logical index as a logical index contains physical indexes that cannot belong to multiple logical indexes.
A BIA index is a special index that is used for SAP NetWeaver 7.0 BI Accelerator. Like the compound index, it consists of several table indexes and metadata.
The table indexes contain the data from an InfoCube. The table indexes are connected using foreign key/key relationships and form a type of star schema. Each table index is either a physical or logical index (depending on its size).
The metadata contains information about which table indexes belong to which BIA index and how to link the table indexes together.
The graphic below shows a BIA index:
As with the compound index, the table indexes can belong to multiple BIA indexes.
A memory index is an index that only exists in the main memory and provides all the TREX standard functions.
TREX uses memory indexes when classifying on the basis of queries. The indexes are automatically generated and deleted by TREX.
A temporary index is an index that only exists in the main memory and only provides an attribute search.
It only contains attributes that can be processed by the attribute engine. It is normally used for the contents of a database table.
Temporary indexes are used as part of high-performance BI to map hierarchies.