Cube size - is this a record?

[garry cook]

I've just come across a service with a 7 dim cube running at 78.6Gb direct from startup. There's another cube operating at 22Gb but that looks a tiddler in comparison.

Pretty sure this is the biggest one I've seen in a production environment and before I start looking at re-engineering, I'm wondering - is this a record?

God bless 64 bit and its ability to make crazy design run (even if "run" is a relative term)

[Wim Gielis]

For me personally, this is a record by far.
For the models that I've been working on until now: I've never come across a model (entire TM1 model running as a service or application) that would not fit in a 8GB RAM server.

In this particular case, you cannot coin the term "design".
Any chance to have a look at it and see what disaster decisions have been made when setting up this cube?

Wim

[garry cook]

In terms of investigating, I'm trying quite hard to back away as fast as I can with my fingers stuck in my ears shouting "Can't hear you!" to be honest

Sadly though, this tactic has rarely worked in the past so guess it's dissection and re-engineering ahoy. Main problem is that I'm actually scared to open it for fear of blowing the server!

[qml]

The biggest models I have heard of and will soon have a chance to look at more closely run at hundreds of gibibytes (200-400 GiB methinks) and the box required to accommodate them has 1TiB of RAM.

The biggest models I have actually worked with are in the low tens of GiB per model which equates to about 10 GiB per cube.

[JDLove]

I think my business partner has a rather chunky model thats maybe a similar size but its dam rare and in his case its a very complex set of calculations and pretty big data sets as well.
I also know he would have designed it very well as much time and thought went into it.

Personally most models I work with are Budgeting and Planning and they never get beyond 10 to 20 GB maybe for the whole model (thats a pretty big model as well).

[garry cook]

The biggest models I have heard of and will soon have a chance to look at more closely run at hundreds of gibibytes (200-400 GiB methinks)

Damn, I thought I'd got a record for finding the most ridiculous cube ever and unless there's a stupid number of cubes in those it turns out it's miles off

[jim wood]

The biggest one I've ever worked with was 29GB. This was due to a product dimension that contained a few hundred thousand members.

[Harvey]

Guilty as charged, I worked on a model that, at it's peak, was around 80GB.

This was an incredibly complex incentive calculation engine that needed to be dynamic and have visible working of the individual calculation steps available. The model is currently around 25GB since I worked on the feeder efficiency, though, and not quite so scary!

I don't have the stats, but I know of a particular telco implementation that got very large. Some of the developers on that project are on the forum, so perhaps they'll weigh in with a new record.

[John Hammond]

Worried about getting your Model out at the TM1 conference?

Then try FEEDERMAX (R) Pills and your RAM is guaranteed to measure up!

Only $99.99 for a pack of 20. Order 200 or more and receive a 600Watt cooling fan for your tower.

Just listen to the testimonials:

Mr W - Brisbane Australia

We had the only 32 bit machine in the datacentre but 2 weeks of FEEDERMAX (R) saw me add 90Gigs to my RAM. Forget about the 3G switch, I now have a phat, 64 bit machine that is the envy of all my colleagues and my assistant tells me she can really feel that double precision default.