I wrote this in 2010, before Clustered Column Store Indexes, SQL Pools (so rule 3 looks a bit odd), and before date data type was mainstream, but the rest of these still look quite
- Thou Shalt use something that exists in your query as a partition key
- Thou shalt not make up a surrogate partition key
- Thou shalt use the partition key as the first key in the clustered index
- Thou shalt honour the partition key in every query
- Thou shalt make sure you do not have any “select top 1 or max on non-partitioned columns”
- Thou shalt keep it simple and share partition schemes
- thou shalt consider filegroup design in partitioning
- thou shalt try to keep a standard partition design across tables within a solution
- thou shalt be aware that 23:997 is before midnight and 23:998 is actually midnight
- thou shalt leave an empty partition at the end of tables
