TM1 Forum

Hi All,

I need to know what is the maximum volume of data TM1 can hold..
Is it directly dependent on RAM..I mean if your cube estimated size is > 10 GB
do you need to have ram of atleast 15 GB..
I want to know if you guys have any idea about this or handled any similar situaton using TM1..
In our project we are bit confused about Infosphere and TM1..
We know infosphere can handle huge volume of data but not sure abt TM1..
If TM1 supports huge volume of data we may even opt for TM1..
Please advice..

Cheers,
Sudheen

Our model is over 9GB, and it seems like someone here mentioned they have 10+. RAM is cheap, so you may as well get extra. In our experience, no matter how much RAM you think you'll need, you'll eventually want more, so buy a box with room to grow. If you haven't bought yet, have your vendor prove the capacity to you-- that's why they're getting the markup

Matt

Hi There,

We have since shrunk our model down but orginally we had a cube that contained every item we sold. This cube was 30GB plus. We now have one cube that is 9GB in size. Saying that, our total model size is 25GB at start up and with use for a week it gets up to 40GB. Our server has 64GB available. We did tend to find that our orginal 30GB cube didn't perform very well. This however wasn't just due to the size of the cube, it was to do with a combination of factors: number of members in our product dimension, the size of the cube and most importantly the speed of the CPU's on the box. If you are looking to use big cubes the best piece of advice I can give you is to use the fastest CPU's you can get. In TM1 each read request gets one CPU thread. This means in a multi core setup the speed and efficiency (cache etc) of the individual core is more important than the number of cores. Number of cores comes in to play when you have a larger number of concurrent users accessing the data.

I hope that helps,

Jim.

Hi Jim,

You 've told: "This means in a multi core setup the speed and efficiency (cache etc) of the individual core is more important than the number of cores"
Tell please, this your personal observation or it somewhere is described

Thanks,
Adic

It ought to be written up, but that may be hard to find.

Fundamentally, in current versions of TM1, calculations are single threaded.

When a cell is changed, the calculated values in that cube and all dependent cubes are thrown away. When someone asks for data, it is recalculated and cached.

_Reading_ data is multithreaded, and _Inputting_ data can be, depending on the version you are using. But not calculations. Fortunately the calculation engine is very very good which is why TM1 is the fastest OLAP server available.

This is why you need the fastest CPUs you can afford, but not very many cores.