Use this syntax when neither input fits in memory. Partition the data by categorical attribute (for example, age range), thereby distributing the records to workers by the specified attribute and reducing the comparison times. If the categorical attribute has n values and each range has same size, the comparison times are reduced to 1/n.
Version 1.1SELECT * FROM IdentityMatch ( ON source_input_table AS a PARTITION BY key ON reference_input_table AS b PARTITION BY key IDColumn ('a.id_column: b.id_column') { NominalMatchColumns ('a.columnX: b.columnY' [,...]) | FuzzyMatchColumns ('a.columnX: b.columnY, match_metric, match_weight [, synonym_file ]' [,...]) } [ Accumulate ('{a|b}.accumulate_column' [,...]')] [ Threshold ('threshold') ] );