Why do cpio/tar/dump/pax all backup to tape painfully

Why do cpio/tar/dump/pax all backup to tape painfully slowly?

Added: 04/17/01

cpio/tar = VERY VERY OLD - never designed for DDS or DLT[TM]

fbackup/ftio = designed for Gb backups and modern tape drives

However, the default for fbackup is 1/2" magtapes, which almost no one
uses anymore. ALWAYS use a config file for fbackup with at least these
options:

blocksperrecord 256
records 32
checkpointfreq 1024
readerprocesses 6
maxretries 5
retrylimit 5000000
maxvoluses 200
filesperfsm 2000

Check the manpage for fbackup for other parameters, including what to do
when you run out of tape (the chgvol parameter).

You need huge block sizes for modern tapes. Try ftio (block size limited
only by hardware), pax (block sizes up to 32k), tar (which is 10k, by
design) or cpio -B which changes from 512 to 5k (but still too small).

cpio, tar, dump, pax and similar utils can NEVER backup large files and
are single processes whereas fbackup can read from 6 different files at
the same time.

Note, the large record size (blocksperrecord), checkpointfreq, and
filesperfsm also contribute to tape overhead.



Home
FAQ