TM1 Forum

Hi all,

We have to redesign TM1 Actual loads process for our client. Currently it take 29 steps to upload data in all the cubes. There are 5 modules for the TM1 system. We have 29 buttons in a websheet and user run one by one each button to complete the load process. We want to make it as short as possible. Where shall we start from?

Some TI Processes are dependent on other while some are not.
Any suggestion is appreciated!!

You're going to have to give a lot more detail, and also ensure your question adheres to the guidelines http://www.tm1forum.com/viewtopic.php?f=3&t=1037

29 sounds like a lot, but it really rather depends on what your loads are doing.

"Client"? Are you a TM1 consultant?

Developer.
The Process has TI's which are updating subsets, loading attributes, loading data, to different cubes for every period(every month) like copying forecast current to forecast prior, loading actuals etc. I also need to know where do we recommend to have staging cubes? Updating subsets and attributes are not happening in the same TI as there are depenedency on another cube data( some cube data is updating attributes) . Also the performance of the system is good so we need to take this point into consideration that while merging TI's performance shouldn't get impacted.

What sort of response are you expecting?

Only you know the existing processes and know the dependencies of the Actuals Load process through all 29 processes.

If there are 29 processes and it works fine and operates at a satisfactory level performancewise then what's the issue? Excessive administration? Are any of these processes scheduled? Do they require manual intervention? Are they passing parameters which can or cannot be automated?

It is such a wide ranging question it could be pulled into any direction. For example why not create one master process with EXECUTEPROCESS for each of your 29 processes.

I'd suggest fulfilling the guidelines as Paul has requested and you might get a better response

It sounds like the problem is the amount of buttons... not the amount of processes. So just create one button that triggers the processes of all the other buttons.

I think thats as much help as this forum can give you.

Thanks for the replies. I can't post all the TI's which are involved in the process as it would be lots.
Just wanted to have an idea like what all factors I should consider if i want to redesign an existing process.

What tests do you have in place to validate that any changes to your TIs have not broken anything?

That is your biggest requirement, ensuring that you have a validation process that you can run to confirm that nothing is broken.

Firs I would create a control cube with the process order and dependencies, this cube stores start and end flags to monitor what is running and what is ready. Then a process checks the cube and triggers those 29 processes in the correct order and following the dependencies to move forward, so you can have parallel processing as many processors you have and the dependencies let you do it.

To trigger multi processes: tm1runti.exe
To make multi process update the control cube: Synchronized

I am going to merge TI processes which belong to same dimension or cube. These processes were created long back and looks scattered, so currently, in a TI we are updating subset and another TI for attributes, then these attributes are getting loaded into cube so some discrepency is there. I will not delete any code but just in place of 3 TI's I will have one TI. I am lanning to convert some TI into year end actvity TI as it updates time dimension for every month from 2012 to 2038, which is not reuired. Regarding consistency, we need to manually assess( I don't see any other option).So I would be able to bring down 29 buttons to 12 as of now since dependency is there.
Also some parameters are getting passed into processes so I can't have 1 master TI calling all other TI. Some manual work is also involved, like converting future projects into active projects, assigning site(its a model for shipbuilding company) to them.

As per the control cube solution, What exactly it would contain? A dimension with all the process names and a dimension with all cubes, another with dimension?

The control cube should have the dimension 1: processes you need to run and dimension 2: "dependency", "start date" and "end date". Each process you run should update "start date" in the Prolog start and "end date" at the Epilog end. This way you can evaluate which processes have being started / completed. You can also create an item "duration" to calculate it.