fregatepallada - Thursday, December 7, 2017 1:17 AM
Heh... "Complete LUNACY". And, yet, the authors of that good document (and it IS well written by those with a high pedigree) say the following in the conclusion of their document.
We did not investigate the behavior of NTFS or SQL Server when multiple writes to multiple objects are interleaved. This may happen if objects are slowly appended to over long periods of time or in multithreaded systems that simultaneously create many objects.
Neither have the authors provided the code that they've actually done their testing with so that one could verify their methods.
Also, and to ask the question, have you actually done a performance comparison where you've methodically applied the scientific method and documented your findings or are you sharing the same anecdotal feelings that the authors of the document are trying to dispel?
--Jeff Moden
Change is inevitable... Change for the better is not.