Performance Issue in FACT Build

[tneogy1]

Hi,

I am devolving a sanity fact build. My build contains the following
1.6 data source sourcing same database
2.Data source query contains group by ,order by,2-3 sub queries
3.Input properties used as 'Merged Record with Duplicate Keys'
4.There are 6 keys set as Dimension and dimension domain size set as maximum for each case.

I am facing performance issue with the build. Build processing stuck showing below message :
Processing Dimension Domains

Can anyone help how to boost up the performance?

Thanks in advance.

[MFGF]

Sounds like it is taking a while to load and cache the dimensions.  What exactly have you set the domain sizes to be on each dimensional build element?  Have you tried running the build with the 'Internal' logging option enabled to see how memory is being utilised?

MF.

[tneogy1]

Thanks for the reply.

Dimension domain size set as 999999999 for each dimension.
Memory settings is as follows:
Max page size(bytes) : 65536
Initial page table size(slots) : 2000000
Initial Hash table size(slots) : 2000000

[MFGF]

999 million? That's a *massive* domain size for each dimension - why so big? This alone could be causing the problems you are seeing.  How many combinations of keys are you likely to find in your incoming fact data? You also appear to have scaled the hash table up by a factor of 10 - is this necessary?  I really hope you are not going to be utilising a page table too - what is the rationale behind setting this to 2 million slots?

How have these options been arrived at?  Have you run the build in check-only mode with internal logging enabled to see how big the domain sizes need to be, the hit rate on the hash table, how full the hash table is getting, whether it is resizing, and whether the page table is being utilised at all?

MF.