Time taken for Exact same view to refresh

[MSidat]

Hi, Wondered if anyone could assist with an issue we are having on one of our Financial Models.

I have opened a view for the first time in a cube at a highly consolidated level and it takes around 20 seconds for it to display. This is as expected for the first time.
Then without anyone changing any of the source data at all I click on recalc in Cube Viewer or open it again after closing the first instance down and it still takes the same amount of time for it to come back, I would have assumed the values would have been cached and as no 'dirty marker' has been set against the view it should come back instantly.

Looking in TM1Top whilst it is doing this i can see that the function is shown as GetViewByHandle and for the 20 seconds or so it cycles through displaying the names of some of the source cubes followed by " - rules".

It is a relatively complex setup with around 7 source cubes feeding into this cube and 2 virtual cubes.

We are using 64bit 9.5.2 FP1.

[jim wood]

Was any data in any of the 7 source cubes updated?

[declanr]

Does this happen with all cube views or just this specific one?

Its quite a rare one to be using but I have seen a certain parameter in action before... Do you have this parameter set in the tm1s.cfg.... UseStargateForRules=F?

It would be worth investigating the VMT and VMM parameters... just play about with them in dev and see what results you get would be my suggestion.

[MSidat]

Was any data in any of the 7 source cubes updated?

Nope, nothings changed at all.

Does this happen with all cube views or just this specific one?

Its quite a rare one to be using but I have seen a certain parameter in action before... Do you have this parameter set in the tm1s.cfg.... UseStargateForRules=F?

It would be worth investigating the VMT and VMM parameters... just play about with them in dev and see what results you get would be my suggestion.

UseStargateForRules is set to True.

VMT/VMM - Never really had to use these before, will have a place with these parameters. Are changes to these dynamic or will a restart of the service be required.

[tomok]

Do you have rules in this cube that are calculating based on attributes and, if so, do you have rules on the attribute cubes?

[mattgoff]

Do you have rules in this cube that are calculating based on attributes and, if so, do you have rules on the attribute cubes?

Is there a rule of thumb related to caching (or lack thereof) of attributes?

Matt

[tomok]

Is there a rule of thumb related to caching (or lack thereof) of attributes?

The rule of thumb is that rule-based text attributes do not cache. Whether that means a view built off rules that reference those same attributes don't cache as well, I don't know. It's worth mentioning in case the OP wants to test that condition.

[mattgoff]

The rule of thumb is that rule-based text attributes do not cache. Whether that means a view built off rules that reference those same attributes don't cache as well, I don't know. It's worth mentioning in case the OP wants to test that condition.

Is that limited to attributes only or all string cells?

[tomok]

Is that limited to attributes only or all string cells?

Haven't done any testing on that so I'm not sure. Would stand to reason that it would be the same for all string cells since there really is no difference.

[MSidat]

Tomok,

Yes I have rules that look for attributes against a time dimension which does have rules in it and also against my Contract dimension which has a rule in it to determine its first parent.

I have managed to remove all the rules in the attributes against the time dimension as these should only be updated once as year.

I am presuming doing an Elpar in my Rule files for my affected Cube will have the same impact as doing an attrs against an attribute which has an elpar rule against it?

PS. I am only using the Elpar attribute in my feeders in this cube and 2 virtual source cubes which undertakes calculations at a higher level.

[Catherine]

It would be worth investigating the VMT and VMM parameters... just play about with them in dev and see what results you get would be my suggestion.

I agree with declanr. I once met the problem, and increasing the VMM parameter solved it.
Have you tried it?
I guess it does not require a restart of the service; as it is a data in a cube, it should be directly in RAM, and taken into account.

[tomok]

I am presuming doing an Elpar in my Rule files for my affected Cube will have the same impact as doing an attrs against an attribute which has an elpar rule against it?

PS. I am only using the Elpar attribute in my feeders in this cube and 2 virtual source cubes which undertakes calculations at a higher level.

As long as the ELPAR rule is not used to populate a string value then you should be OK. While it can be very convenient to use rules to populate attributes, especially for the time dimension, don't do it if you are using those attributes to write other rules against. I always populate them with TI, or via a spreasheet (like the time dimension which only needs updating once a year).