The present invention provides a data migration technique enabling a required performance to be satisfied as closely as possible. A server obtains an IOPS measurement value of a plurality of business APs and an IOPS measurement value of a plurality of logical volumes serving as the access destinations of the respective business APs, and on the basis of the various obtained IOPS measurement values, selects a source sub-candidate from a plurality of VOLs. The server identifies a parity group to which the selected source sub-candidate belongs from a plurality of parity groups, sets the two or more VOLs belonging to the identified parity group as two or more source candidates, and selects a plurality of target candidates from a plurality of other VOLs. The server then selects the source candidate/target candidate pair that is forecast to perform migration most favorably and/or forecast to produce the most favorable migration result, and performs data migration from the source to the target constituting the selected pair.