CDR with multiple legs

Hi All,

Currently we have a challenge to model CDR DW for tracking the voice calls inside big enterprise which uses different telecommunication subsystem.

Each individual call may include a couple of legs (couple of CDR records)
It is important to keep information about legs because it influences the charges and customer want to have ability to track the flow to optimize the routing approches

I'd like you to suggest approches to model the fact tables.

Is it worth to have 2 fact tables, one with 1 leg granularity and one with the 1 call one?
Does it make sense to try to implement accumulative snapshot fact table? but the possible flows are quite vary to have fixed step milestones...

Is it worth to have call granularity fact table with multy valued dimension? (using leg group bridge table)

Thank you for your thoughts in advance,
Andriy

in general does someone know which approach could be used to design accumulate snapshot fact table with the dynamic workflow?

You want to capture the data at the lowest grain. Whether you create an aggregate table or view for summarized data is a simply a question of performance. An accumulating snapshot of CDR's is probably not feasible due to the quantity of updates and potential table size.