Reporting (no metadata) cube load causing waiting - why?

[jyoung66]

Situation: 9.5.2 “development” environment.
A chore that had several TIs that simply wrote data (no metadata) from read / write cubes with rules to a read only reporting cube was causing other threads to “Wait”.

Why might this be happening I asked? Parallel Interaction should have meant no waiting.

After speaking to several guru mates of mine, I concluded that the most likely cause was cube dependencies.

Here’s the hypothesis: cube dependencies were not being re-established and rules were being changed and added to often, meaning that the dependencies were also being re-established but at the time the chores were running, causing server wide locking.

What do you think of my hypothesis? Can you suggest other things that may have caused the threads to wait?

[EvgenyT]

Hi there,


Read the link below:

http://www-01.ibm.com/support/docview.w ... wg21460595 chapter: TM1 9.5.2 Operations that cause Object Locking (and potential performance delays).

It should give you some insight into your problem.

Your guru mates are probably correct. As the guideline above states:

Feeder statements having data dependent cube references or rules files containing ATTRN or ATTRS functions will not establish Cube Dependencies during server load

So when you run your TI it probably tries to establish those dependencies, since they haven't been established during the server load. Also can you please post the feeders for the affected cube?

What do you guys think?



Thanks

ET

[jyoung66]

Thanks ET

Yes, I was sent this by one of my mates, this and some discussion led me to make my assumption.

I thought I would post it here as well to see what the wider community thinks

Thanks again for your reply

[EvgenyT]

You are welcome.

You can try to establish dependencies using AddCubeDependency(BaseCube, DependentCube) function in TI straight after the server start-up.

Have a go at it and let us know if it proves performance for you.

Thank you

ET

[jyoung66]

Unfortunately I can't test this, as this was a problem I had that the start of the year on a project that is now complete

Re-establishing at server start up wouldn't do it though would it, as there would need to be a re-run of the AddCubeDependency(BaseCube, DependentCube) function after each change to the rules if I understand correctly? Changes in rules would blow away the dependancies?

[EvgenyT]

Re-establishing at server start up wouldn't do it though would it, as there would need to be a re-run of the AddCubeDependency(BaseCube, DependentCube) function after each change to the rules if I understand correctly? Changes in rules would blow away the dependancies?

Yes dependencies are blown away when rule complies i.e. rule deleted, modified or dimension changes happened (*which cause rules to compile*). I wouldn't suspect that rules would change very often? If there are metadata changes in the affected cubes' dimensions then you could run AddCubeDependency after those processes?

Thanks

ET

[jyoung66]

In a PRODUCTION environement rules dont change often, but this was in a UAT or SAT environment and changes were more frequent, this is where I saw the behaviour

[EvgenyT]

Perhaps you can use it next time