Palo Community Forum

Hello sebast,

Of course there is a relation between the input data and the size of the cube.
It depands on witch version of Palo/Jedox you are using?
In the palo community 3.2 version is a lower compression rate, than in the Jedox 3.3 verison.

Also it depands on the bit length of your values/lines. I estimate that you need near 2 GB, with the CE 3.2 Version.
In your case, i would go ensure and take a computer with 4 GB RAM.

Regards.

Hello sebast,

of course the length of each cell is important.
The size of a cell depands on, what the cell is defined. If it is a numeric, e.g.: int (i think palo save an int value as a float), then palo reserved as much bits as the size of a float is defined.
If it is a text, then it is as long as the text is.
The limit of a dimension is 4 million elements (3.3 Jedox).

Just as an example:
If you have 7.000.000 lines of information and in each line, there are 2 columns (price, pieces), than you double your cube size to 16.000.000 cells.
Also when you have a dimension with 10 measures, you get 10 multiple 16.000.000 cells, and so on.
Also the structure and logic is important. If you have a "complex" tree structure for example, it will also increases your cube size.

I don't know if there is a formular and i wouldn't say/guess anything about the cube size, but if you find one, please let me know.

Regards.

Hi,

thats maybe for the simpliest structure correct in a way, but i don't believe.
Like I said above, everything is important, the structure, the logic, the number of your measures and so on.
I think that palo is calculating that in an other way.
I opened a support ticket (because now i want to know it exactly)

Regards.

Hello sebast,

here is the offiziell answer of jedox support, how to calculate the cube size:

Quoted

This is a very rough and unofficial estimation of the memory consumption.

up to 80 bytes per element in flat dimensions + string length
up to 250 bytes per element in "real-life" dimension + string length
you can find details here http://bugs.palo.net/mantis/view.php?id=8596#c33092

up to 16-20 bytes per numeric value in a cube
the same + string length for a string value

if you have markers, e.g. [budget, feb] = 2 * [[budget, jan]]
then for each value with coordinate [budget, jan, ...] additional 16-20 bytes

I have no time to "test" it, i hope it help you!

Regards.