Separate output from pg_dump into multiple files (dump.####.<name>) Intended to maximize compression of de-dup'ing backup system Theory: By separating table data into individual files, (1) de-dup'ing of unchanged tables is very high, unaffected by other tables; (2) tables with new (appended) records can still be highly compressed because older records (positioned earlier in the dump) are de-dup'ed; (3) only new tables and tables with changed rows have lower compression with less de-dup'ing Files are numbered to maintain order for reconstruction First file (dump.0000.pre-schema) contains first part of schema information Next files contain table data, in same order as pg_dump output Last file (dump.####.post-schema) contains last part of schema information Use: pg_dump | split-pg-dump.pl To reconstruct complete dump (for restoring, etc) cat dump.* > dump.txt