Impact of field size on sorter performance in Datastage

For this test i created two almost identical streams. The only difference between the streams is the field length of 10 NVarchar fields. The first stream has fields with a length of 25. The second stream has fields with length 250.

Both streams wite data to /dev/null to limit the influence of IO to a minimum.

Stream 1 writes data with 28,000 rows per second and stream 2 writes with 7,267 rows per second so stream 1 is 385% faster.

The conlusion is simple. If you're sorting your data, make sure that all field lengths are set to the minimum possible.