Storing spectrometer data with 1024 channels is a problem in SQL Server unless you pivot (which has too much baggage to be useful at that scale IMO) since the storage engine limits us to 1024 columns per table. The problem can be handled quite nicely however using a SQLCLR Type while providing solid performance and type-safety. SQLCLR types open up many possibilities and help us overcome many of the limitations L' Eomot Inversé alluded to with storing scientific data (many readings per entity per time) in a relational store.