As I mentioned, if you cannot use a datetime to distingush rows because they arrive too fast for the resolution of the clock, so more than one row gets the same value, you'll have to depend upon some other form of a "uniquifier" to distingush the rows.
By far, the simplest is to use an identity column. The values are assigned by the system in a monotonicaly increasing order so you can use this value to distingush two rows with the same datetime value. Disregading some unusual circumstances, you can depend upon the fact that a row with a smaller identity value arrived "before" a row with a larger value.
Custom Apps, Inc.