Circular Cube Dependency

[foogy]

Hi there,

When analyzing a TM1 model which some consultants left behind at our company, I accidentially found out that there is a circular cube dependency, i.e. cube A depends on cube B and vice versa. Now I'm curious what this means in terms of locking, caching, or other side effects.

We have 9.5.2 FP2 with parallel interaction turned on, if this makes any difference.

From my understanding of cube dependencies, this situation would at least destroy any benefits related to caching.
, so this might be a reason for some of the performance issues we observed. However, from time to time, the system stops responding completely without any further entry in the logs, which looks like a deadlock situation.

What do you think? Does TM1 handle such circular dependencies somehow or did I find a huge weakness in the design of this model?

Thanks. Regards,
foogy

[lotsaram]

I wouldn't actually think a circular cube dependency was that unusual. Although it could represent a model weakness more than likey it doesn't and is quite normal with no ill effects. For example in many planning models there may be a driver input cube and a main calculation cube (due to different dimensionality). The main calc cube will obviously be dependent on the input cube but it's pretty common to pull through summary stats into the driver cube from the calc cube so the dependency goes both ways. From a caching perspective any data entry in either cube invalidates the cache but this does not mean caching is ineffective. Where there is no change to entry data then cell values and views will be cached perfectly well. There is no "destruction" of any caching benefits through this.

TM1 handles this just fine. If there are other issues you might need to look in more detail elsewhere.