1

Topic: Re: olap out of memory

1 wrote:

Kind night!
out of memory at cube processing
How to understand, on what there is no storage and how it is possible to optimize?
Thanks

I suggest to go stage by stage. Version SSAS. The size of operative storage on the server. Presence of large measurements. And the most important thing - a cube that what, MOLAP or SSAS Tabular?

2

Re: Re: olap out of memory

1;
If MOLAP, the most simple check that at you is not present a deficit operative - to watch during large measurements of files khstore . Perhaps, you beforehand do not process measurement generally, and at ProcessFull a cube they start to be processed beforehand ProcessFull, that is a problem not cubed, and in number of elements in measurements which are connected to it...

3

Re: Re: olap out of memory

Andy_OLAP;
the Link about files khstore , in the previous message overshot.

4

Re: Re: olap out of memory

1;
Plus the good link with speculations from Dzhessa Orosha that is necessary for measurements - a snowflake or a star . In general, begin with ProcessFull the cube measurements, one behind another, and  the storage expenditure.

5

Re: Re: olap out of memory

1 wrote:

it is passed...
Thanks, fall on a measure (distinct count) 300 000 000
16 
Model molap
mssql 2012

Put only for this measurement of 48 Gb. And for a cube as a whole still. So 64 is a minimum from which you will work. 16 is  even not ridiculously...

6

Re: Re: olap out of memory

Andy_OLAP;
Oh-twist, overshot, speech not about measurement. ProcessStructure a cube, then ProcessData groups of measures Distinct Count transits? Then ProcessIndex transits? Falls on ProcessFull groups of measures or on ProcessFull a cube when business reaches this group of measures?
The maximum size of measurement on which key it is considered distinct count, what?

7

Re: Re: olap out of memory

1 wrote:

it is passed...
Thanks, fall on a measure (distinct count) 300 000 000
16 
Model molap
mssql 2012

And I so assume, what on partite Partitions (section) you did not begin to break group of measures, with one blow decided to process 300 million?

8

Re: Re: olap out of memory

1 wrote:

it is passed...
Yes, but
1. I think 12  should suffice on 300
2. distinct it is impossible to beat on partition

1. If large measurements are not present. And if Aggregation Design did not fasten what is offered a wizard from MS, there totally ludicrous, it is better to remove in emptiness and once again to try.
2. It is possible. We admit, at you check turns. And an amount of checks the counter distinct count on ID the check. Cut both groups of measures on section on 1 year either on 1 quarter or on 1 month, fasten competently Slice - and process one section for another if storages do not suffice.