becaus the increment by 1 is handled vproc local.
Identity columns are not guarantee a sequence - they only quarantee unique ids.
Each vproc request a range of numbers he can assign next. The size of the range is controlled via a dbs controll field - and don't even think of setting this to 1 as you would not have a MPP process any longer.
So assumumg the dbs controll field is stateing 1000 and you have 2 vprocs.
so the first vproc will start with 1 and the second with 1001.
If all numbers of the range assigned to the vproc are used he is requesting a new range (next would be 2000) etc.
So if you add only 100 rows per vproc you would see (in perferct world) values between 1 and 100 and 1001 and 1100.
becaus the increment by 1 is handled vproc local.
Identity columns are not guarantee a sequence - they only quarantee unique ids.
Each vproc request a range of numbers he can assign next. The size of the range is controlled via a dbs controll field - and don't even think of setting this to 1 as you would not have a MPP process any longer.
So assumumg the dbs controll field is stateing 1000 and you have 2 vprocs.
so the first vproc will start with 1 and the second with 1001.
If all numbers of the range assigned to the vproc are used he is requesting a new range (next would be 2000) etc.
So if you add only 100 rows per vproc you would see (in perferct world) values between 1 and 100 and 1001 and 1100.