> So the in-place update is a vital part of the program, because a range of values > of column col should be stored together on the disk. Is there any way to do such > a in-place update without generating much extra overhead? >
Although in-place update is not possible, can he not use partitioning to at least try to store the data together based on the value of column col?