TM1 Forum

Dear all,
I got one problem in our production.
There's a cube with large amount data, when we startup the TM1 server (no demand load). And our user double click the big cube(just name it A), it will show the default View.
My question is: will TM1 load all the data of the cube or just load the data which the default view need only?

"Load" is not the proper terminology I don't think as the user cannot load data by opening a view. I think you mean "retrieve" or "calculate". When a view is opened TM1 only retrieves and/or calculates what's defined in the view. If that view has a lot of high level consolidated elements in it then it can take a long time to open the first time.

wuruimao wrote:Dear all,
I got one problem in our production.
There's a cube with large amount data, when we startup the TM1 server (no demand load) (sic). And our user double click the big cube(just name it A), it will show the default View.
My question is: will TM1 load all the data of the cube or just load the data which the default view need only?

I'm assuming that you've made a typo there and actually meant "On Demand Load" or, more correctly given the menu option, "Load on Demand".

Bear in mind that this doesn't work with some cubes where inter-cube rules and feeders are involved, but let's leave that aside.

The first time you select a view on a Load On Demand cube, the entire cube loads into memory, not just your view. It has to be this way because of course it's the only way the cube can be sure that it has all of the values that it needs to do any calculations. (Not to mention feeders if rules, even ones internal to the cube, are involved.)

That means that you're simply postponing the load from the server startup, to the first time that the server is queried. The first query will therefore take a pretty fair whack of time, depending on the size of the cube.

The only time this option is really useful is for cubes which are rarely queried. You save yourself a little load time and memory capacity if your bet that no-one will query the cube pays off in that session. If the cube will always be queried in a session, Load On Demand merely postpones the inevitable and there's really little point using it.

wuruimao wrote:My question is: will TM1 load all the data of the cube or just load the data which the default view need only?

Just to go a little further on what Tomok and Alan already said, it isn't clear if you actually understand that the way TM1 works is that all cube data already is loaded into RAM, that is simply he way TM1 is (taking aside whether load on demand is active for the cube). When a user opens a view the entire cube (leaf level) data is already in memory to load the view the server performs the calculations necessary for the view (possibly with a little more on the side for row and column dimensions depending on a few configuration settings) but in principle that's it, no more, no less.

tomok wrote:"Load" is not the proper terminology I don't think as the user cannot load data by opening a view. I think you mean "retrieve" or "calculate". When a view is opened TM1 only retrieves and/or calculates what's defined in the view. If that view has a lot of high level consolidated elements in it then it can take a long time to open the first time.

Hi tomok,

I know what u r talking about. For example, assuming a cube is 100G,large amount of data.
And in it's default view, for every dimension, I just choose one element, N element.
So the view has only one line record.

My question is : whne the cube is not yet loaded to the memory, we double click to open the default view. Will the TM1 load 100G data , or just load the data the default view need( only one line record, so it will take less time? )?

Thanks..

Alan Kirk wrote:

wuruimao wrote:Dear all,
I got one problem in our production.
There's a cube with large amount data, when we startup the TM1 server (no demand load) (sic). And our user double click the big cube(just name it A), it will show the default View.
My question is: will TM1 load all the data of the cube or just load the data which the default view need only?

I'm assuming that you've made a typo there and actually meant "On Demand Load" or, more correctly given the menu option, "Load on Demand".

Bear in mind that this doesn't work with some cubes where inter-cube rules and feeders are involved, but let's leave that aside.

The first time you select a view on a Load On Demand cube, the entire cube loads into memory, not just your view. It has to be this way because of course it's the only way the cube can be sure that it has all of the values that it needs to do any calculations. (Not to mention feeders if rules, even ones internal to the cube, are involved.)

That means that you're simply postponing the load from the server startup, to the first time that the server is queried. The first query will therefore take a pretty fair whack of time, depending on the size of the cube.

The only time this option is really useful is for cubes which are rarely queried. You save yourself a little load time and memory capacity if your bet that no-one will query the cube pays off in that session. If the cube will always be queried in a session, Load On Demand merely postpones the inevitable and there's really little point using it.

Thanks very much Alan.