Database Reference
In-Depth Information
You have more choices when you process a dimension. You can specify
Pro-
cessFull
that will do a full reprocessing of the dimension. You can also specify
ProcessUpdate
that will reread all the data, but will only add updated information
to the dimension store.
ProcessUpdate
can actually be slower than
ProcessFull
for an object, the reason is that the dimension has to be checked if it contains the
latest version and this takes time. But, as
ProcessUpdate
does not clear the object
as part of the processing, it will affect fewer objects and it will not require reprocess-
ing of the other objects.
With dimensions, cubes, and partitions which will be discussed shortly, you have the
option to do
ProcessData
and
ProcessIndexes
as well. There can be a bene-
fit of executing these commands separately instead of doing
ProcessFull
.
Pro-
cessData
just reads the data from the source but does not create indexes, the ob-
ject will be available for querying after
ProcessData
has been executed, but you
will not have the same performance as you will have when the indexes have been
built.
ProcessIndexes
just creates the indexes separately and requires that
Pro-
cessData
has been executed.
Tip
More information about how to tune processing and how it can be monitored
can be found in the Microsoft SQL Server Analysis Services Multidimensional
Performance and Operations Guide available at
http://download.microsoft.com/
crosoft%20SQL%20Server%20Analysis%20Services%20Multidimensional%20Performance%20and%20Operations%20Guide.pdf
.
More information about processing can be read in the Analysis Services 2005
Processing Architecture whitepaper available at
http://msdn.microsoft.com/en-
US/library/ms345142(v=SQL.90).aspx
(even though it is written for 2005 it is still
relevant for SQL Server 2012).
Sometimes processing even on partition level (covered later) is hard to fit into the
batch window available. In that case there is an option that you can use and that
is having a separate processing server and then synchronizing the query server
with a processing server. You can read more about this option in the following
article from the SQLCAT team:
Search WWH ::
Custom Search