SSAS-Info.com - Microsoft SQL Server Analysis Services (SSAS) 2005/2008 resource hub
SSAS Tips
Filter     Display # 
# Article Title
1 Process only one partition so you can browse the cube
2 Using unformatted cell values it may improve query performance
3 NULL default measure might improve querying performance
4 Split large partitions, with more than 20 million rows or that are over 250 MB in size, into ...
5 Set the Slice property on partitions with ROLAP storage mode or with proactive caching settings...
6 Remove aggregation designs that are not used by any partitions
7 Limit the number of aggregation designs in a measure group to three
8 For partitions, do not use proactive caching with a polling query but no processing query
9 Consolidate partitions for measure groups with 5 or more partitions that have less than 2 million...
10 Avoid specifying ROLAP storage mode or enabling proactive caching with the OnlineMode property ...
11 Do not build more that 500 aggregations for a partition
12 Design aggregations for partitions with 500000 or more rows
13 The names of attribute relationships should match the names of the related attributes
14 Create hierarchies in non-parent child dimensions
15 Avoid visible attribute hierarchies for attributes used as levels in user-defined hierarchies
16 Set the AttributeHierarchyVisible property for the key attribute in a parent-child dimension to Fals
17 For dimensions, do not use proactive caching with a polling query but no processing query
18 Either disable the attribute hierarchy or set the AttributeHierarchyOptimizedState property to ...
19 Do not ignore duplicate key errors. Change the KeyDuplicate property of the error configuration ...
20 Define the default member for non-aggregatable attributes in either the DefaultMember property or ..
21 Define attribute relationships as 'Rigid' where appropriate
22 Change the UnknownMember property of dimensions from Hidden to None
23 Avoid redundant attribute relationships because these relationships may prevent data from being ...
24 Avoid large parent-child dimensions that have 500000 members for the key attribute
25 Avoid diamond-shape relationships not covered by hierarchies
26 Avoid defining non-aggregatable atributes in a parent-child dimension with the exception of the pare
27 Use numeric key columns for attributes with 500000 or more members
28 Design hierarchies such that lower levels have more members than higher levels
29 Avoid specifying more than one non-aggregatable attributes per dimension
30 Avoid creating hierarchies where attribute relationships do not exists between one or more levels
31 Define a time dimension
32 Avoid having 10 or more dimensions with a single attribute. Consider unifying them
33 A database should have no more than 3 parent-child dimensions that have ...
34 Define partitions for measure groups
35 Avoid non-materialized reference dimensions
36 Avoid having 15 or more measure groups in a cube
37 Avoid defining measure groups with zero dimensional overlap
38 Avoid cubes with a single dimension
39 Include the default measure of a perspective in the perspective
40 Consider setting dimensions that contain measure groups with semi-additive measures to MOLAP
41 Break distinct count measures into separate measure groups
42 Avoid defining many-to-many relationships when either the intermediate measure group ...
43 Avoid creating measure grops that have the same dimensionality and granularity
44 If you can, list in SELECT statement real measures first and calculated measures after
45 Do not forget to set properly attribute relationship type (rigid or flexible).
46 SP2 has a lot of performance improvements for processing and querying. Upgrade ASAP.