Memory Usage by TM1 Cubes

[ppillai]

Hi,
I am looking into Performance issues we are facing with TM1 application, We have got two archive cubes which are set as Load on Demand, but it looks like when the TM1 Server is started these two cubes are loading in memory. These cubes are not accessed by any other cubes but has got some rules defined on these cubes. What my understanding was that if the property of the cube is set as Load on Demand , these cubes will not load in memory until it is been accessed.

Please see the TM1 Server log lines, which shows that Feeders are calculated on these cubes, does this mean that cubes are loaded in memory while service is started?

3340 INFO 2010-06-19 16:27:54,687 TM1.Server Computing feeders for cube "CashFlow Archive"
3340 INFO 2010-06-19 16:28:33,531 TM1.Server Done computing feeders for cube "CashFlow Archive"

3340 INFO 2010-06-19 16:29:13,953 TM1.Server Computing feeders for cube "ProfitLoss Archive"
3340 INFO 2010-06-19 16:29:38,156 TM1.Server Done computing feeders for cube "ProfitLoss Archive"

We are using TM1 9.4 MR1version.

Thanks and Regards
Praseeja

[Alan Kirk]

I am looking into Performance issues we are facing with TM1 application, We have got two archive cubes which are set as Load on Demand, but it looks like when the TM1 Server is started these two cubes are loading in memory. These cubes are not accessed by any other cubes but has got some rules defined on these cubes. What my understanding was that if the property of the cube is set as Load on Demand , these cubes will not load in memory until it is been accessed.

Provided that there are no rules or feeders involved, yes. However...

Please see the TM1 Server log lines, which shows that Feeders are calculated on these cubes, does this mean that cubes are loaded in memory while service is started?

3340 INFO 2010-06-19 16:27:54,687 TM1.Server Computing feeders for cube "CashFlow Archive"
3340 INFO 2010-06-19 16:28:33,531 TM1.Server Done computing feeders for cube "CashFlow Archive"

3340 INFO 2010-06-19 16:29:13,953 TM1.Server Computing feeders for cube "ProfitLoss Archive"
3340 INFO 2010-06-19 16:29:38,156 TM1.Server Done computing feeders for cube "ProfitLoss Archive"

We are using TM1 9.4 MR1version.

As you've found, if a cube has feeders it has to load regardless. I'm not completely sure whether the server is able to distinguish between feeders which are completely internal or those which feed other cubes; obviously if there are any of the latter, the cube must load to ensure that the other cube is fed. However I suspect that it doesn't make the distinction and needs to load the rules before it can know whether any of the feeders are external. (I haven't tested this, though.)

If these are purely archive cubes one thing that you may want to consider is making a copy of the cube with no rules at all and feeding the calculated values into that cube via TI. It depends on how long that would take to do (though you'd only need to do it once) and it also depends on how many potential calculated values there are. However it may save some memory.

[ppillai]

Thanks for your reply. All the Rules are internal the to the cubes, as you mentioned we are storing most of the calculated values, the rules which we need to define are specific for the Consolidated level, hence we can not use the TI process to calculate and store these, I presume if we need to calculate the consolidated levels differently we need to define these rules, otherwise normal TM1 consolidation will apply is this correct?