string loading with TI

Post Reply
nicola531
Posts: 43
Joined: Thu Oct 22, 2009 7:58 am
OLAP Product: TM1
Version: 9.5.1
Excel Version: 2007

string loading with TI

Post by nicola531 » Fri Feb 08, 2013 3:03 pm

Hello everybody,

I'm wirting to understand what is your position related to this fact:string transfer via TurobIntegrator.

OLAP technologies are not so well structured, as far as I know, to manage strings.
In my opinion TM1 is definitely the case.
In my case I try to transfer attributes values from one dimension to another, but when it is about strings the TI loading is very slow: we are speaking of 400 records per senconds and my dimensions contains more or less 100 K elements.

I would like to understand if you have any best practise to imporve this speed.
I tried:

- normal turbo integrator process with ATTRPUS function and CELLPUTS function
- Repplication of the Attribute cube, but in the repplication options I cannot change the }atributecube measure dimension

Any other ideas?

Let me know.

Thanks in advance.
N.

User avatar
ioscat
Regular Participant
Posts: 209
Joined: Tue Jul 10, 2012 8:26 am
OLAP Product: Contributor
Version: 9.5.2 10.1.1 10.2
Excel Version: 07+10+13
Contact:

Re: string loading with TI

Post by ioscat » Fri Feb 08, 2013 4:12 pm

try to switch off rule recalculation while loading data
no idea will it actually help

User avatar
paulsimon
MVP
Posts: 718
Joined: Sat Sep 03, 2011 11:10 pm
OLAP Product: TM1
Version: PA 2.0.5
Excel Version: 2016
Contact:

Re: string loading with TI

Post by paulsimon » Fri Feb 08, 2013 10:54 pm

Hi Nicola

I am guessing that your background is in ROLAP rather than MOLAP. Are you aware that you can use the same dimension in more than one cube. That means that there is no need to transfer attributes between two dimensions.

Assuming that the two dimensions are different in some way, another approach to using TI to physically copy the attributes is to use rules in the }ElementAttributes_<dimension> cube to pull the values across.

Are you turning off Cube Logging on the destination }ElementAttribute_<dimension> cube?

You don't say how you are doing the transfer. If you are reading from a View on the }ElementAttributes cube, then make sure that you have SkipZeroes ticked on the source View.

Does the source cube have any rules. If so then skip calculations.

The Skip consolidations does not work correctly with Strings, since strings don't consolidate. If you only need base level attributes then use a subset on the dimension.

Another approach is to use the source of the Dimension Subset, and then use an Attrs to get the relevant Attributes, and AttrPuts to store them. You will need to experiment to see whether a source of the View or Subset gives the best performance.

As the previous person said, removing not so much rules, but feeders, on the destination }ElementAttributes cube may prevent feeders from firing as each value is written.

Another potential problem might be dynamic subsets defined on the destination dimension that are re-evaluating. However, in generaly these will only evaluate on request.

Are there any possible concurrency issues - are others accessing the dimension as it is being written?

Are you transferring attributes from a dimension on one server to a dimension on a different server? With replication it is possible to rename the dimension. However, it is not something I have used a great deal. I would generally go for export to a file and then import. ODBO might be a possibility but again this is not something I have used much.

Regards

Paul Simon

nicola531 wrote:Hello everybody,

I'm wirting to understand what is your position related to this fact:string transfer via TurobIntegrator.

OLAP technologies are not so well structured, as far as I know, to manage strings.
In my opinion TM1 is definitely the case.
In my case I try to transfer attributes values from one dimension to another, but when it is about strings the TI loading is very slow: we are speaking of 400 records per senconds and my dimensions contains more or less 100 K elements.

I would like to understand if you have any best practise to imporve this speed.
I tried:

- normal turbo integrator process with ATTRPUS function and CELLPUTS function
- Repplication of the Attribute cube, but in the repplication options I cannot change the }atributecube measure dimension

Any other ideas?

Let me know.

Thanks in advance.
N.

Post Reply