question about feeders

[Wirt]

Hello,
Explain to me, please, why the rule with feeders have a long time to save.
for example - Rule has 5 settlements and 10 feeders, Internal 5 and 5 to other cubes, save about 30 minutes, without feeder calculations quickly save. And that in all cubes.
What could be the reason?
Thanks!

[Edward Stuart]

Read this, http://www.ibm.com/developerworks/data/ ... ge620.html and if you have further questions post them here

[Wirt]

Thanks for the link, I have read, and I have a question - In this article write that You can see how much memory is used for calculations feeders ( in cube }StatsByCube ), but it does not say how much it should be used. ( for example - "Look for cubes with particularly large values ... " )
I looked, and I have a value of more than 5 million in two cubes, in other cubes ( 3 ) have a value of about 1 million, basically all values up to a million ( from 200 thousand to 700 thousand )
So, approximately how much memory should be used? Can you prompt me?
Thank you!

[Edward Stuart]

Method 6 – Use the Performance Monitor
This method will not specifically show which FEEDERS are over feeding, but it will give an idea of where to start looking. Often TM1 Developers are presented with a finished or partially finished model which is running slow and is using up a lot of memory.
Start the performance monitor in TM1 Architect by right-clicking on the TM1 Server name and selecting “Start Performance Monitor” from the menu as shown in Figure 12. Confirm that “Display Control Objects” as found under the View menu option has been enabled.

Confirm that “Display Control Objects” as found under the View menu option has been enabled.
Open the “}StatsByCube” system cube in IBM Cognos TM1 Cube Viewer and notice it contains the following information on FEEDERS as shown in Figure 13. The “Feeders” line shows the number of fed cells and memory used by the FEEDERS under the columns titled “Number of Fed Cells” and “Memory Used for Feeders” respectively. Look for cubes with particularly large values under these columns. Use an “overfeeds” cube as described earlier in Method 5 to determine whether or not any calculations are being over-fed.

The values being referred to are within the }StatsByCube cube, not your actual cubes

Whilst it is probably possible to anticipate exactly how much memory Feeders 'should' take per cube my focus is always that the values are fed appropriately. Do your values hold when Zero Suppressed, can you Check Feeders and Trace Feeders?

[Wirt]

"Do your values hold when Zero Suppressed, can you Check Feeders and Trace Feeders?"

Yes, All these paras executed. When I check FEEDERS, the first thing I look thet the data displayed in consolidation, If the cube does not have consolidation, I check with Zero Suppressed.
The problem is not that data is not feed, the problem is that for a long time saved rule.
Maybe the reason is that the cubes are large ( an average of 10 dim. ) ?

Here is an example of calculation in a cube, which uses memory more than 5 million, the exact number is - 5,628,690,432. Why so much?
In some cubes More feeders, and they use less memory.
( real name of dim. I changed )

SKIPCHECK;
[] = N: IF (
ATTRS ( 'Dim1', !Dim1, 'attribute1' )@=
ATTRS ( 'Dim2', !Dim2, 'attribute2' ),
DB('Cube',!dim3,!dim4,!dim5,!dim6,!dim7,!dim8,!dim9,!dim10,!dim11), STET );

FEEDERS;
['element']=>DB('Cube 2',. . . , ' element');

Thanks!

[tomok]

Your example makes absolutely zero sense. How do you expect anyone to provide you any assistance without a real life example? The time it takes to save a rule is directly correlated with how many cells must be fed since every time you save a rule the feeders must be re-evaluated. What you need to figure out is why you have so many cells fed. Maybe it's correct, as in you really have that many intersections in the cube where the rule is populating a value, or maybe it's not, as in you have the condition commonly referred to as overfeeding. No one can provide that answer to you with what you have provided so far.