• I understand the clustered index doesn't need to be unique, however the column chosen for partitioning (InsertedLogId) is very dense (only about 300 values so far) so wouldn't make a good clustered index on its own.

    The reason for choosing InsertedLogId as the partitioning column is because A) it is static, never updated, and B) it is often used in WHERE clauses and could hopefully allow partition elimination.

    Does anyone agree this leads to needing a compound clustered index? If so, any suggestions on which fields, and what order they should go in?