Modelling minute-level granularity
Posted: Tue Sep 19, 2006 11:53 am
We have a datamart that consists of a factless fact table that keeps track of the statuses of contracts which are loaded daily (incremental). This has allowed users to prepare various kinds of reports. Now, there is a strong voice for extending the mart to allow analyis of response time that it took for contract to go from one status to another.
Now, since a contract may undergo a number of status changes in a day, the granularity of TIME will have to be 'minute' and not 'day' like we currently have. This would also mean that the all the dimensions tie up to the fact on some 'Minute_ID' and not 'Day_ID'. Further, the dimensions will not be conformed to our existing BUS.
However, We seem to have no option, but to create this as an isolated datamart satisfying one particular requirement. Is it possible that the requirement be fulfilled without the new model, but by making some changes in the existing daily granule model itself?
Thanks
Now, since a contract may undergo a number of status changes in a day, the granularity of TIME will have to be 'minute' and not 'day' like we currently have. This would also mean that the all the dimensions tie up to the fact on some 'Minute_ID' and not 'Day_ID'. Further, the dimensions will not be conformed to our existing BUS.
However, We seem to have no option, but to create this as an isolated datamart satisfying one particular requirement. Is it possible that the requirement be fulfilled without the new model, but by making some changes in the existing daily granule model itself?
Thanks