TM1 Hardware Requirements

[spudmusic]

Good morning colleagues!

Can you tell me something about TM1 hardware requirements.
TM1 9.4 MR1 is the most interesting thing.
I've seen software requirements on IBM website, but there's no any word about hardware.

Thanks!

Regards,
Alex

[Martin Ryan]

Morning Alex,

RAM RAM RAM!

TM1 stores all its data in RAM, so if your model is going to be quite big, or progress to be quite big you want to get your hands on as much RAM as possible.

Otherwise, any modern server is perfectly acceptable to run a standard TM1 model.

In order to be more specific, can you tell us more about what is it that you intend to do with TM1?

Cheers,
Martin

[spudmusic]

Martin, thanks a lot for your answer!

Because TM1 is now a Cognos and IBM product we're going to implement it and offer to our customers.
And hardware requirements to run a common TM1 model are needed to give them to our potential clients.
I understand that hardware depends on the size of the model and so on, but i'm sure common requirements exist

[Martin Erlmoser]

1-2 quadcores perform very nice

[Martin Ryan]

It all depends on what your customer is after. As a general rule of thumb though, if you are going to put large amounts of data through the system and/or reporting is going to be the main focus, then RAM is crucial and they'll want to look at going 64 bit if possible.

If there is going to be a lot of rules and what-if type analysis, then a powerful processor is also important.

Personally I wouldn't want to make a sweeping statement about exactly what hardware requirements are necessary for a "common" application because every application differs so much, but someone else might be braver.

Martin

[George Regateiro]

Personally I wouldn't want to make a sweeping statement about exactly what hardware requirements are necessary for a "common" application because every application differs so much, but someone else might be braver.Martin

I for one will not be brave, but just give you a history of our TM1 model evolution.

Our first model very simple P&L Balance Sheet system only took a gig of RAM and worked fine on a virtual server. The next application which was alot of forecasting has now grown to over 5 gigs in a relatively short amount of time (this model will come nowhere near a virtual server). So though it might be disconcerting to say there are no real hard guidlines, it is a true statement.

Like was already stated barring a very compelling reason 64 bit should be your starting point and just load up on the RAM. In my opinion the processor capacity of today's servers should be sufficient for most models, going light on the RAM will be the killer of most models.

[mattgoff]

Multi-threaded loading, at least on 9.1 does crazy things to RAM requirements too. I hadn't really noticed this until I recently noticed that production (4-core, 3 load threads) is pushing 6gigs but dev (1-core) is down around 2.5gigs. If I remember correctly, this was something they addressed in 9.4, but I haven't had time to even think about an upgrade recently....

Matt