I have a sales database that has become too large. I need to break it up into pieces so that the physical files are manageable. It makes clear sense to break it up according to sales date. Each year will have its records put into a separate physical file via filegroups. I've done this on other databases, and it was straightforward.
In this database, there is a TransactionHeader table, which has the sales date, and a unique varchar(20) TRANID field that uniquely identifies the transaction. There are then 8 different supporting tables that use the TRANID as both their primary key, and a foreign key into the TransactionHeader table. Makes sense... you only need to store the transaction header info once.
It's easy for me to create a partition function/scheme for the TransactionHeader table, since I can use the TransactionDate to know where...