Database Reference
In-Depth Information
however, that the Analysis Services engine must scan fact tables (or views)
to build aggregations. Reading millions of data rows will be time-consuming,
even for building aggregations, hence your processing will be slow. You can't
simply run
ProcessIndexes
on partitions after running
ProcessUpdate
on dimensions while using ROLAP; you must fully reprocess partitions, which
in this case will mean scanning fact tables specifically to build the aggrega-
tions.
• If you're processing multiple objects in parallel and see a smaller number
of SQL statements sent to the relational source than you expect, based on
degree of parallelism, experiment with
BufferMemoryLimit
and
BufferRe-
cordLimit
options. SSAS could overestimate the amount of memory needed
for processing each object and, therefore, throttle the number of objects pro-
cessed. Lowering the value of the mentioned settings can help improve the
processing performance by working on more objects in parallel. Also, exam-
ine the
Maximum Number of Connections
data source property. Ensure
you allow enough connections to your relational data source, but not too
many. For example, if you're processing 64 partitions in parallel, but you only
allow 10 connections to the data source, you will see that only up to 10 quer-
ies will run in parallel and the rest of them will queue up. If your relational
database can indeed handle 64 parallel queries, bump up the
Data Source
property to, maximum of 64 connections.
Search WWH ::
Custom Search