Daniel Bartley on 20 Jan 2017 19:48:45
I have a source system with lots of uniqueidentifiers. I do ETL to strategically flatten the data model and change the data type of relationship keys and bitint to int64 (database is third party and overzealous for our purposes.
Could SSAS engine detect that some fields are unique identifier / hash and skip the dictionary step to reduce bloat? Ie. the processing engine encodes the hash as int64 and throws away the gibberish or dumps it in a redundant temp file.