Cube View takes an hour to appear

[Waterlily]

Cube view takes a long long time, 1 hour to come up. Something is wrong, but do not know why. Cube is small, 150k, it has 8 dimensions ( dim1 has 7 elements, Scenario has 12 elements, Company has 3 elements, Machine Group has 12 elements, Year has 10 elements, Period has 18 elements, Measure has 20 elements, Part has 10,000 elements). I have a view that has all dimensions and 20 Parts, it comes up no problem. If I try to show all Total All Parts, then the system will come up with message that says "Cube View Processing - Building Cube View (Stop Building View)". This will take over 30 minutes and still not come up with view. If I pick "Stop Building View", it will have "TM1 Error : CubeViewer - :Error creating view array:". There are plenty of RAM, 60G, with 8G dedicated to tm1sd.exe and 24G available.

I did try to reorder the cube dimension, using System suggested method. I have not used Stargate VMM and VMT yet. Something is fundamentally not well, can you give me a hand? Thanks.

[Alan Kirk]

Cube view takes a long long time, 1 hour to come up. Something is wrong, but do not know why. Cube is small, 150k, it has 8 dimensions ( dim1 has 7 elements, Scenario has 12 elements, Company has 3 elements, Machine Group has 12 elements, Year has 10 elements, Period has 18 elements, Measure has 20 elements, Part has 10,000 elements). I have a view that has all dimensions and 20 Parts, it comes up no problem. If I try to show all Total All Parts, then the system will come up with message that says "Cube View Processing - Building Cube View (Stop Building View)". This will take over 30 minutes and still not come up with view. If I pick "Stop Building View", it will have "TM1 Error : CubeViewer - :Error creating view array:". There are plenty of RAM, 60G, with 8G dedicated to tm1sd.exe and 24G available.

I did try to reorder the cube dimension, using System suggested method. I have not used Stargate VMM and VMT yet. Something is fundamentally not well, can you give me a hand? Thanks.

Does it have any rules? If so, are you using SkipCheck and feeders?

[mce]

Try calculating the view after enabling zero suppression.

[tomok]

The ONLY reason a cube of this size would take that long to load is because you've got some serious calculations going on and you forgot to use SKIPCHECK and then wrote FEEDERS.

[Gregor Koch]

Hi

If the zero suppression doesn't fix your problem than you might have a Feeder problem. But there are a few considerations to take into account.
Really depends on wether you indeed have rules on that cube and what your view does.
With what you are saying about the size of the cube I doubt that you actually have a lot of Feeders set, as they need memory as well.

You mentioned your view has "all dimensions and 20 Parts" and that it works fine.

I assume that your view shows certain elements out of all the other dimension apart from the Part dimension and not all elements.
When you change to "all Total All Parts" is that a consolidation or are you trying to show or the 10.000 parts on N-Level?

Note, even if you have no rules at all but try to create a non zero suppressed view which tries to display 200.000.000 zeroes, TM1 obviously will have a hard time doing that.

Cheers

[Waterlily]

"If the zero suppression doesn't fix your problem than you might have a Feeder problem. But there are a few considerations to take into account.Really depends on wether you indeed have rules on that cube and what your view does.
With what you are saying about the size of the cube I doubt that you actually have a lot of Feeders set, as they need memory as well.

You mentioned your view has "all dimensions and 20 Parts" and that it works fine.

I assume that your view shows certain elements out of all the other dimension apart from the Part dimension and not all elements.
When you change to "all Total All Parts" is that a consolidation or are you trying to show or the 10.000 parts on N-Level?

Note, even if you have no rules at all but try to create a non zero suppressed view which tries to display 200.000.000 zeroes, TM1 obviously will have a hard time doing that."

REPLY =>

1. There is zero suppression, SKIPCHECK and FEEDERS in rule
2. I inherit the cube from someone who left, the cube does have some crazy and serious conditional rules, total 160 lines, connecting 4 cubes, perhaps I need to review the rules...
3. The view does have zero suppression, so I will not be displaying 200,000,000 zeroes.

I recently changed cfg "MaximumCubeLoadThreads" to properly use the multi cpu cores, it seems to improve the situation. Before the change, system will always hang cause only 1/8 of the cpu is available, then I have to restart service. After service is restarted, the RAM of tm1sd.exe is always brought down to 5-6 G, then I have to reload all the views, after 2 hours and the views are loaded, tm1sd.exe then goes up to 9-10G, then the system will hang due to some other reasons, and the tm1sd.exe will go down again to 5-6 G. The cycle will repeat.

Now with the change of "MaximumCubeLoadThreads", the system seems to be much more stable, the tm1sd.exe remains at 9-10 G even after the restart, thus I do not have to manually reload the views again. I cannot really explain all these. Will report back if the system really stabilize.

[Gregor Koch]

Hi

"MaximumCubeLoadThreads" does not improve how you retrieve a view. It is a setting purely for the way you start up your server.
TM1 will always only use one Thread/Core to retrieve the values for a view for one user.
So I'd say this is coincidental rather than a real fix for your problem.

I should have mentioned this earlier:
Could you start the performance monitor and check (in }statscube) how many feeders are set against the cube?
Just to sort out the Feeder question. Still not convinced that it is the only issue, at least for that specific cube.

You could also check whether you actually get a result for your ‘Total Parts’ consolidation at all if you pick one N- Level element each out of all other dimensions. In that case you would be consolidating 10.000 cells which are usually a piece of cake for TM1.
If that takes a long time, your rules are crazy and they need an overhaul.
It’s also possible that you have a circular reference in your rules.

Cheers

[bihints.com]

2. I inherit the cube from someone who left, the cube does have some crazy and serious conditional rules, total 160 lines
...
Now with the change of "MaximumCubeLoadThreads", the system seems to be much more stable, the tm1sd.exe remains at 9-10 G even after the restart, thus I do not have to manually reload the views again. I cannot really explain all these. Will report back if the system really stabilize.

Did you mean you also have conditional feeders? These do not mix well with that parameter as explained in the manual.
http://publib.boulder.ibm.com/infocente ... 503FE.html

If this is the case that could explain these conditional feeders were previously causing trouble and have been skipped with that multi-thread load?

[jstrygner]

If the zero suppression doesn't fix your problem
(...)
Note, even if you have no rules at all but try to create a non zero suppressed view which tries to display 200.000.000 zeroes, TM1 obviously will have a hard time doing that.

Guys,
I was just hit by the "Error creating view array" problem.
I have a larger cube (20 Dimensions, two biggest are 600 000 and 1 000 000 elements) and want to display a view with 600 000 rows and 30 columns.
For now I have only zeroes in my view (it will get some data later).
I have rules for the cube, but just a few with a SKIPCHECK and without FEEDERS (not needed for consolidations in my case). Anyway those rules don't "touch" cells I want to see in my view.
Zero suppression does NOT fix my problem.

Does it mean I need to build smaller arrays while looking for data in the cube (those work fine)?
Or maybe there are some settings I could set to make such views work?

Regards.

Server: Windows Server 2008, version 9.5.1
Client: Windows XP, version 9.5.1 HF9, Excel 2003

[tomok]

I was just hit by the "Error creating view array" problem.
I have a larger cube (20 Dimensions, two biggest are 600 000 and 1 000 000 elements) and want to display a view with 600 000 rows and 30 columns.
For now I have only zeroes in my view (it will get some data later).
I have rules for the cube, but just a few with a SKIPCHECK and without FEEDERS (not needed for consolidations in my case). Anyway those rules don't "touch" cells I want to see in my view.
Zero suppression does NOT fix my problem.

Does it mean I need to build smaller arrays while looking for data in the cube (those work fine)?
Or maybe there are some settings I could set to make such views work?

You're not going to be able to open a view that large in the Cube Viewer, not even close. I don't know what the maximum size is going to be because it depends on your environment. I would try starting with a view that has 10,000 rows and go up from there by 10,000 until it crashes, That will then be your maximum.