I use ADO (AppendChunk) in a C++ program to write binary
data file into a table column of type image.

The image files have almost the size of 1 MBytes. When i
measured the time to save 1000 data files (before and
after update method) i am very surprised that the time
varies between 100 ms (96%) and 30000 ms (4%). I expected
that the time could not be changed so strongly.

On the other side when i read these binary data the time
is almost constant.

How can i understand this result ? I think when SQL
Server needs to expand the database it needs more time ???

Any hint would be great.