1

Topic: SSAS: AttributeRelationships RIGIT vs FLEXIBLE - query Performance

About that is whether a difference in productivity for MDX requests on hierarchies, if dispute came into them communication type rigid or floppy here at us.
That that influences processing it is known, me productivity of requests on the big measurements with several levels interests.

2

Re: SSAS: AttributeRelationships RIGIT vs FLEXIBLE - query Performance

Can not absolutely in a subject I will answer, but:
I do practically always Rigid communications since at me Process Data and Process Index are spaced apart on separate steps.
And if values of attributes of measurement change, it is necessary to fulfill Process Index, and to read from sources the data in the facts of groups of measures (Process Data) it is not necessary.
Well and in measurements of value of attributes can change often and for a considerable quantity of elements:
That on a relational source of clients  in other segments many requests in other statuses passed, apparently, on absolutely ancient records something there corrected

3

Re: SSAS: AttributeRelationships RIGIT vs FLEXIBLE - query Performance

Greg Galloway says that the difference is not present:
https://social.msdn.microsoft.com/Forum … isservices
Indexes and  it is necessary for floppy communications, of course, with Process Index to staticize.

4

Re: SSAS: AttributeRelationships RIGIT vs FLEXIBLE - query Performance

Yuri Abele;
Not absolutely in a subject. Somehow analyzed productivity of requests cubed with a considerable quantity of partitions and M2M communications in a stage,  was on time. For partitions were exposed Partition Slice according to elements of hierarchy of time dimension. The intermediate requests were analyzed at selection on M2M to measurements, especially on the indirect.
And so, while communications in DimDate were flexible - jambs at selection of the intermediate partitions M2M was much, that is in certain cases there was a scanning of all partitions. After transfer on rigid communications - empty scanning stopped.
The environment of experiments - SSAS 2012.

5

Re: SSAS: AttributeRelationships RIGIT vs FLEXIBLE - query Performance

Ferdipux;
Well, not the fact that the reason in it. There can be it there was a banal advancing reading.

6

Re: SSAS: AttributeRelationships RIGIT vs FLEXIBLE - query Performance

Alex_496 wrote:

Can not absolutely in a subject I will answer, but:
I do practically always Rigid communications since at me Process Data and Process Index are spaced apart on separate steps.
And if values of attributes of measurement change, it is necessary to fulfill Process Index, and to read from sources the data in the facts of groups of measures (Process Data) it is not necessary.
Well and in measurements of value of attributes can change often and for a considerable quantity of elements:
That on a relational source of clients  in other segments many requests in other statuses passed, apparently, on absolutely ancient records something there corrected

How it works? After all in case of tight couplings full-processing is required after the element changed the position in hierarchy. In case of the big cube it is fraught with that in the beginning of the working day cubes will be not updated.
In my judgement tight couplings can be used only in calendar measurement, or in other cases of similar measurements.

7

Re: SSAS: AttributeRelationships RIGIT vs FLEXIBLE - query Performance

The critic;
If at the Author of measurement not in ten millions elements that to worry for full processing measurements. If natural hierarchies in 4-5 levels with good coefficient of the relation of level to the previous level - generally a class.
now  not to surprise 1 operative storage.
By the way, who can prompts as still to accelerate process index the big measurements

8

Re: SSAS: AttributeRelationships RIGIT vs FLEXIBLE - query Performance

Alex_496 wrote:

the Critic;
If at the Author of measurement not in ten millions elements that to worry for full processing measurements. If natural hierarchies in 4-5 levels with good coefficient of the relation of level to the previous level - generally a class.
now  not to surprise 1 operative storage.
By the way, who can prompts as still to accelerate process index the big measurements

So measurement small, and the facts a heap. Besides in the course of handling it is necessary separate to do a branch in which it will be necessary  all cubic
On the big measurements I researched a question - at me all rested against productivity on a kernel. Probably, in the code of processing SSAS there is a piece, which not .

9

Re: SSAS: AttributeRelationships RIGIT vs FLEXIBLE - query Performance

wrote:

Ferdipux;
Well, not the fact that the reason in it. There can be it there was a banal advancing reading.

Perhaps and so. But such full scanning arose often and on request repetition.
Upon - passage on rigid in time dimension helped communication with our case on cascade M2M requests aloud, even users noted it.

10

Re: SSAS: AttributeRelationships RIGIT vs FLEXIBLE - query Performance

The critic;
Something I do not understand what for to do full-process of groups of measures when change  between attributes of measurements.
Process Data - it is added new  records to the last partitions, and here Process Index - that yes.
Also it is that yes (recalculation of aggregates) at me is more least on time than remaining phases Process.

11

Re: SSAS: AttributeRelationships RIGIT vs FLEXIBLE - query Performance

wrote:

. How it works? After all in case of tight couplings full-processing is required after the element changed the position in hierarchy. In case of the big cube it is fraught with that in the beginning of the working day cubes will be not updated...

I will support Alex_496 and I will a little disagree with the Critic - not mandatory to do ProcessFull, the relationship type for problem measurements can be changed on a live cube, it does not change a state {Processed/Unprocessed} measurements/measures.

12

Re: SSAS: AttributeRelationships RIGIT vs FLEXIBLE - query Performance

. Though Alex_496 for some reason speaks about measures while speech is faster about hierarchies.

13

Re: SSAS: AttributeRelationships RIGIT vs FLEXIBLE - query Performance

Alex_496 wrote:

something I do not understand what for to do full-process of groups of measures when change  between attributes of measurements.
Process Data - it is added new  records to the last partitions, and here Process Index - that yes.
Also it is that yes (recalculation of aggregates) at me is more least on time than remaining phases Process.

because at tight couplings (and changes in them from the measurement data) - measurement processing falls out with an error (since in algorithm there is no operation of reset of aggregations for this relationship type), is 2 decisions:
() -  (Full) corresponding MG
() - to change communication type, Process (Update on measurement, Index on MG) to return communications (if there is a necessity)