TM1 Forum

Hi Team,

I may be asking a silly question.
TM1 is an in-memory OLAP, as per the definition all the required data gets loaded in to the memory and all calculations gets performed from there. So it makes things faster.
I understand that data getting loaded in to the memory (RAM) of your 'TM1 Server'. Is there any dependency on the RAM of your 'TM1 Client' on the performance ?

Regards
Sree

I'm guessing you've not gone through the documentation that comes with the software? The answer to your question is no, but possibly may be. For example if the user is viewing large amounts of data then it will use more memory, but that has nothing to do with performance, that's to do with model design and training. Again you really need to go through the documentation to understand the TM1 architecture. Not only will it answer your question but others that may come up related to this one.

sreesuku wrote: I understand that data getting loaded in to the memory (RAM) of your 'TM1 Server'. Is there any dependency on the RAM of your 'TM1 Client' on the performance ?

Generally not as Jim mentioned. However you should still ensure that the client machine has sufficient memory to operate without needing to use virtual memory. With some notebooks, for instance, still being sold with a whopping 2 Gig of RAM you can start to run out of physical memory. The resulting disk swapping will cause a performance impact on all applications, not just TM1. 4GB is usually enough, again depending on how many other tasks you're running.

I'm guessing you've not gone through the documentation that comes with the software? The answer to your question is no, but possibly may be.

In which documentation could you have found this answer?

Thanks for the replies

So one agreed point is - When a request comes, the data gets loaded in to the 'Server RAM' itself. But 'Client RAM' also should be at a considerable level so that it can perform the TM1 tasks just like any other tasks. Is my understanding correct ?

I have another point. How about when using TM1 Web ? Where users are not installing any TM1 client components ?

mvaspal wrote:

I'm guessing you've not gone through the documentation that comes with the software? The answer to your question is no, but possibly may be.

In which documentation could you have found this answer?

The operations guide covers the TM1 architecture. May be not as well as it could be but it's there.

sreesuku wrote:Is my understanding correct ?

Yes

sreesuku wrote:I have another point. How about when using TM1 Web ? Where users are not installing any TM1 client components ?

The information is still processed server side. As for memory on the client it depends on your version. Below 10.1 is .Net based so some information processed on the client side. I've not really looked at the new java version in 10.2 yet so I can't comment on that.

Memory counts in every PC application so the more the better. Most of the processing in TM1 is server side so that's where the RAM needs are the greatest. IBM has server sizing guidelines to help you figure out how much. When it comes to the client, there are no guidelines. One thing I do know is that dimensions get downloaded to the client (as far as I know this has not changed), so if you have a model with extremely large dimensions, like 100,000 elements plus, then you need to be on the higher limits for client RAM. Skimping by on 4GB might cause problems. I would go for 8GB or more on your client machines. This is for Architect and Perspectives. This is not applicable to TM1Web.

As tomok has suggested when you are using perspectives you want to go as chunky as you can get. If you are using tm1web most end users will possibly not use perspectives so you just need to worry about developers PCs. The client site I am working on has laptops with 4gigs and I've noticed much slower interaction than I'd like / am used to.
My personal laptop now has 32gigs so I can also run servers on it and my old laptop was 8... the 8 was ace for using connected to a server but running any meaty servers locally and I wouldn't go lower than my current setup.

Your other option is to remote onto the server if possible and work with perspectives on there (or a server.)