Designing Real-Time Cubes - Professional SQL Server Analysis Services 2005 with MDX - page 962

Figure 18-17
Billions and Billions of Records
The American astronomer Carl Sagan seemed fond of pondering the number
of stars in a galaxy or cluster of galaxies; well, these days it is not a stretch to
have similarly mind-boggling numbers of records in your transactional reposit-
ories. For reference, the Milky Way contains about 200 billion stars. Anyway,
you don't have to ponder how to deal with them because we have a recom-
mended solution for you right here. If you have a large number of fact data
(on the order of billions of rows), ask yourself if the dimension data does not
change much as compared to the fact data that changes regularly. If this is
the case, building the cache might take a disproportionate amount of time
since the dimension and fact data needs to be updated. Typically you will
have the fact data split across hundreds of partitions. However if the fact data
is changing frequently, and if you do need real-time access of the data, then
the cache needs to updated frequently and needs to be merged with existing
partitions for the new fact data and Analysis Services needs to aggregate the
data from multiple partitions to the end users.