Is there an issue with UPDATE STATISTICS tablename with FULLSCAN when
there exists a large number of _WA_Sys_(columnname)_(guid) auto
statistics attached to the table?? If one does this command against
an extremely large table and then performs a DBCC SHOW_STATISTICS
(tablename, _WA_Sys_(columnname)_(guid)) for each of these auto
statistics in order, the Updated time differences between each of
these auto statistics is significantly higher than the clustered index
or any of the non-clustered indexes.
If I run a script to only run UPDATE STATISTICS tablename indexname
WITH FULLSCAN, I get extremely reasonable times, say fewer than 5
mins. However an UPDATE STATISTICS tablename autostatsname WITH
FULLSCAN can easily take 2 to 3 times to perform. Is some internal
processing occurring that causes such an extreme overhead??