I have found the checksum to be useful in some cases and unreliable in some. If the data set if distinct then the checksum function (of all rows stands good), but in another case i had to get 27 Million rows from DB2 with no primary key, no uniqueness on the table, so i thought of using checksum to compare the rows in a staging environment. Not only did i have the overhead of creating the checksum, i also had to compare it + the checksum gives me a integer value based on the row uniqueness, where i had duplicates and everything was getting messed up..... so in short i would think the usage totally depends on the case.