Hi Folks,we are doing a database Migration that involves a conversion of non-Unicode data to Unicode. The process involves the conversion of every non-Unicode table in the database totalling around 2TB of data.The first stage of the Migration was to restore the database to a new Server which went without a hitch. Because the restore took around 4 hours I decided to take a snapshot of the database to err on the side of caution. What happened next surprised me: The conversion, which had been tested several times over although without the snapshots was running between 10-15 times slower than the practice runs on the same Server.When I looked a Little deeper I discovered in the error logs that there was massive contention in the TEMPDB, something we hadn't seen during practice. Furthermore, the database has 10 files and trace flag 1118 is also turned on. The question at this Point was "why was only one tempdb being used and not all 10"? When I used the Task Manager to see what activity was occuring, there was around 4x as much activity as in the source database! I can't understand why....When I removed the snapshot the database went back to its normal very fast state with no further interaction needed from me....Can anyone please explain this phenomenon and why a snapshot can so adversely affect database Performance?Many thanks in advance...Regards,Kev
↧