Microsoft Confidential
We look at the world...
with our own eyes ...
And no
matter what
our
perspective
is...
We understand that people
play the most important role
People interact, communicate and
relate each other every day ...
They relate in a one to many
relationships ...
On a one to one
relationships ...
And in a many-to-many
relationships ...
With all the pros and cons ...
for each one of them
So for the SSAS …
... Cube perspective of the
world ...... 
The way relationships are build…
is extremelly important
Specially when we are talking
about M2M relationships
Data measure
group
Intermediate
dimension
Intermediate
measure group
M2M dimension
14
Query
IN
MEMORY
INNER
JOIN
By M2M Dim
16
As the data size of the records in the join
increases
The performance of the run-time join in
SSAS suffers
IN
MEMORY
INNER
JOIN
17
Query performance in SSAS is intrinsically
and linearly tied to the size of the data
being joined between
Data and
Intermediate measure groups
18
There are 3 optimization techniques:
Defining aggregations
Partitioning
Matrix relationship
19
1st
What is an aggregation?
Pre-calculated summary of data that SSAS
uses to enhance query performance
Moves the calculation to the processing phase
Summarizes measures by a combination of
dimension attributes
20
Aggregations
Work by reducing the number of records that
the storage engine needs to scan from disk in
order to satisfy a query
Provides a significant benefit only if the size of
the aggregation is significantly smaller than the
size of the original table
21
Where to define aggregations in a M2M
scenario?
22
Data measure group aggregations for
M2M queries
Include in the aggregation the granularity
attribute of all dimensions that join with the
intermediate measure group
Do not include the attribute from the M2M
dimension (occurs at query time as part of the
run-time join)
23
Benefits
The benefit is directly related to the size of this
aggregation compared to the size of the data
measure group
An aggregation whose size is >=1/3 the size of
the facts themselves is not considered to be
useful
24
Intermediate measure group aggregations
for M2M queries
Include in the aggregation the granularity
attribute of the dimensions in the
intermediate measure group that relates to
the data measure group
The attribute in the dimension that you wish to
aggregate
25
Benefits
The benefit is directly related to the resulting
size of the aggregation compared to the size of
intermediate measure group
26
2nd
Partitioning
Enables SSAS to retrieve data from only a
portion of data in a measure group
Parallelize data retrieval when data must be
retrieved from multiple partitions
27
Where to define partitions in a M2M
scenario?
28
Data measure group
The data measure group should generally be
partitioned by the same dimension attribute
members as the intermediate measure group
29
Intermediate measure group
You must relate the partitioning dimensions to
the intermediate measure
Benefit
Reduces the data size used by the run-time join
between intermediate and data measure groups
When the M2M query can be resolved from only a
few (or one) partitions
30
If the M2M query must be resolved by
retrieving data from many or all partitions
The technique provides little or no value and
may even increase response times
31
Design patern
The intermediate measure group must
Be partitioned by one or more common dimensions
used in queries
Be related to each dimension that is used for
partitioning
The intermediate fact table must
Contain the dimension surrogate key
That is used for the measure group partitioning
To relate with the dimension
32
3rd
Increases M2M query performance by
reducing the size of the run-time join
Uses a process of compression to eliminate
unnecessary repetitiveness in the intermediate
fact table
Collapses the size of the intermediate fact table
33
Two phases:
Compression and creation of the matrix key
Implementing the matrix optimization
34
Creates a compressed intermediate fact
table by taking the following steps:
Identify common dimension member
combinations in the intermediate fact table
Each set of common dimension member
combinations is assigned a surrogate key
(matrix key)
Repeated combinations are eliminated
35
36
Requires changes to aspects of the M2M
relationship at the following levels:
The relational data warehouse implementation
level
The cube design level
The ETL implementation level
37
38
Matrix dimension table creation
39
Requires an intermediate fact table that
relates the M2M dimensions to the base
fact through the matrix key
Will have the matrix key column and one or
more dimension key columns
40
Add the Matrix Key column to the base fact
table
41
Works in the same way that any M2M
dimension relationship
The difference is
Shared dimension between the base measure
group and the intermediate measure group is
the matrix dimension, rather than the fact
dimension
42
In Data Source View, add two tables for
Matrix dimension
New intermediate fact table
43
Add a new SSAS database dimension
based on the matrix dimension table
Add this dimension to the cube that
contains the base measure group
44
Add a new SSAS measure group to act as
the intermediate measure group for
resolving the M2M relationships
Remove the existing
Sales Reasons
measure group
45
Define the Dimension Usage to associate
the dimension tables to the measure groups
Both the base and intermediate measure
group must have a direct relationship to the
matrix dimension
46
The intermediate measure group also needs
a direct relationship to the dimension or
dimensions that participate in the M2M
relationship
The dimensions referenced in the second
point must also be related to the base
measure group
47
The ETL involves
Updating the relational data warehouse tables
with new matrix relationships
Associating the fact tables to the matrix
dimensions
48
The general guidelines are:
Adding new key concatenation strings to the
matrix dimension table for new records being
added to the fact table
Adding the matrix dimension key associated
with each transaction ID to the base fact table
Adding the key of the matrix dimension with
each M2M dimension key record
49
Questions and
Answers
50