TI Performance

[deepakjain2020]

Hi All,

Need some knowledge from your end and correct my understanding regarding TI Process.

We are having a TI Process which calls another process in Metadata section.

Source TI process is having a Data Source as "Cube View". Variables are passed as a parameter to another process.
Process which is called is having Five parameters and parameters are passed against all the records in datasource and it is calling another process for every single record.

What my understanding is "As the process is passing parameters against all the records and calling other process it is taking lot of time to complete".

Will like to know is that correct?
Please share your views on what else could be a reason for process to take a lot of time for completion.

Regards,
Deepak Jain

[tomok]

I don't know how you would expect anyone to be able to help you given that you've chosen to ignore the principles in the Request for Assistance Guidelines. That being said, your assumption that the Process_1 is going to execute Process_2 once for every record in the data source of Process_1 is correct. Therefore, if you have 1,000,000 records in the original source then then Process_2 will be executed that many times. I can't give you any more than that.

[Wim Gielis]

You should restrict the source view to really understand what is going on. For example, a view containing 5 cells as the Data Source.
Then you should be able to see 5 ExecuteProcess statements in the TM1 Message Log.

Do you really need to execute the process in the Metadata tab? Can't it be done in the Prolog or Epilog tab (which of course has consequences in terms of timing).
Do you need to execute the other process BEFORE going to the next value in the data source?

[deepakjain2020]

Hi Tomok / Wim Gielis,

Apologies for that.

Process1:

Datasource: Cube View.
Metadata tab has a exceuteprocess(process2)
Parameters are passed from exisitng dimension values.

Process 2:
Datasource: None
5 Parameters are created and the parameters are fetched from process1.

Process2 will insert elements into dimensions.

As it was already exisitng in the metadata, what I felt is that it might be the reason for completion of process1 taking lot of time, is that right?

As part of solution, it would be ok if we take resultset of process1 in .csv file and then take the datasource as a file in process2 and then insert the elements into Dimension.
I have tried taking into .csv file and it is giving appropriate result.

I am trying to figure out and will look to include those statements within the single process and check the solution.

If you have any alternate ideas please share.


Regards,
Deepak Jain

[lotsaram]

The question is what is the 2nd process doing??

[deepakjain2020]

Hi Lotsaram,

As mentioned in my previous post Process2 will insert elements into Dimensions.
If any more info is required, please let me know.

Regards,
Deepak Jain

[lotsaram]

Hi Lotsaram,

As mentioned in my previous post Process2 will insert elements into Dimensions.
If any more info is required, please let me know.

Regards,
Deepak Jain

If you had provided enough information I wouldn't have asked the question. Does the second process have a data source? If so how big is it? Or does the 2nd process execute entirely on the prolog from the parameters? (In which case why have a separate process at all?)

Why do you insist on making it so difficult for other people to help you?

[Gabor]

Problem is, the MetaData tab processes each line of the data source.
If you insert the data on this tab, the TI is working on a dim copy and there is only 1 dim compilation after the MetaData tab is finished, before the Data tab processing starts.
On the other hand, if you call an extra process from MetaData tab, I guess you do it for each line of source data, which requires a dim to be changed. So you may call it many times and each time the closing of the extra process will cause a dim compile.
Depending on the dim size this may take a few seconds.
So your first choice should be to think of a reduction of dim compilations, that can save a lot of time.

[deepakjain2020]

Hi Lotsaram,

Apologies if my replies are giving you a pain. I will re-phrase it again with respect to process2 specifically.

Process2:

Datasource : None
Variables: None
Advanced Tab
1. Parameters: There are Five parameters
2. Prolog: Statements are with respect to parameters which inserts data into Dimension.
3. Metadata: Nothing is written
4. Data: Nothing is written
5. Epilog: Few statements which make amendments in one of the cube.

Hope I am able to provide required information .
If any more information is required please let me know.

Regards,
Deepak Jain

[deepakjain2020]

Hi Gabor,

Thanks for sharing appropriate information.
It gives me some clarity with my understanding of why TI process is taking lot of time.

Regards,
Deepak Jain

[tomok]

So the 2nd process has no data source, it justs inserts members into a dimension and then writes data to a cube? Why are you doing this instead of just inserting the members in the metadata tab of the 1st process and then writing to the cube in the Data tab of the 1st process? Doesn't sound like you need this 2nd process at all. The only reason I can think of for needing a second process would be to have a separate data source from the one used in the first.