Databases Reference
In-Depth Information
Use Incremental Processing when Appropriate
Often, data changes in the relational data source. These changes could
be due to new rows being added to existing tables or updates on existing
rows. If you have set up the storage mode for dimensions and parti-
tions as ROLAP, you will be retrieving the data from the relational data
source for queries sent by users. In situations where the result set has
been cached on Analysis Services due to a previous query, Analysis Ser-
vices will not be fetching data from the relational database by default.
You can force Analysis Services to always fetch data from the relational
data source by using ROLAP (see
Chapter 18
for more details). However,
if the storage mode for all your cubes and dimensions is MOLAP, Ana-
lysis Services serves queries only from the processed MOLAP data that
resides on Analysis Services. Future updates to relational tables will not
be available to end users unless you update the MOLAP data on Analysis
Services.
You have several ways of updating the data on Analysis Services. You
can do a full process of the corresponding dimensions and/or cubes that
need to be refreshed due to changes in the corresponding relational
tables. There are several processing options to optimize your processing
needs. Process Incremental and Process Add are options that help you
process the dimensions and partitions so that they are updated with the
new data on the relational data source as necessary. Not all the data in
the partitions or dimensions gets updated during these operations; only
data that changed since the last round of processing will be refreshed.
During an incremental process of partitions Analysis Services retrieves the new
data from the relational data source and adds it to a temporary partition. Aggreg-
ations are then created for this new data, and finally the temporary partition is
merged to the existing partition. As soon as the data is merged it is available for
querying and you will see the new data reflected in user queries. Because Analys-
is Services only retrieves the new data added to relational tables, the incremental
processing option for partitions helps you to process the partition quickly as com-
pared to full process. However, if your partition is to be processed frequently due
to data changes in the relational data source, we recommend you do a full process
on that partition periodically (not often), because full processing will sort and optim-
ize the data storage of multidimensional data. Similar to the defragment technique
in file storage, sorting the data on disk will improve query performance.
Search WWH ::
Custom Search