Dimension and Partition Processing - Microsoft SQL Server 2008 Analysis Services - page 386

Cube Processing

Analysis Services stores the data of the cube in partitions. When you process the cube, you
load data into all its partitions. Partitions can have one of three storage modes:

. ROLAP —Analysis Services doesn't read any data from the relational database during

processing. It only checks the consistency of the partition's metadata.

. MOLAP —Analysis Services loads data, and builds indexes and aggregations.

. HOLAP (hybrid online analytical processing) —Analysis Services performs all oper-

ations done for the ROLAP partition and builds MOLAP aggregations.

Each partition can be processed independently; therefore, you can process partitions in
parallel. When you process a database, a cube, or a measure group, Analysis Services
schedules the processing of all the partitions contained by those objects. Based on the
resources of the system, Analysis Services attempts to schedule partition processing in
parallel. Analysis Services processes a partition in two stages, as shown in Figure 21.4.

Process
Data Job

Read Data
Job

Process Data
Job

Write Data
Job

Process
Partition Job

Build
Indexes And
Aggregations Job

Segment 1
Index and Aggregation
Job

Segment 2
Index and Aggregation
Job

…

Segment N
Index and Aggregation
Job

FIGURE 21.4

Analysis Services processes a partition in two stages.

. Process Data —Reads data and stores it in the partition files

. Process Indexes and Aggregations —Builds indexes and aggregations