While investigating an unrelated issue on our backup server, I came across an interesting discussion about gzip vs. bzip2. I was surprised to read on how much slower bzip2 is. I even tested it on our server. And as expected, I saw the huge difference.
$ du -sh home 819M home $ time tar czf test.tar.gz home real 3m29.741s user 1m4.026s sys 0m5.629s $ time tar cjf test.tar.bz2 home real 11m38.751s user 6m19.259s sys 0m7.237s $ ll test.tar* -rw-r--r-- 1 leonid users 365987716 2010-06-29 13:08 test.tar.bz2 -rw-r--r-- 1 leonid users 390655750 2010-06-29 12:56 test.tar.gz
For such a small difference in size, the compression time difference is huge! Of course, I should play with more parameters, repeat the tests several times, and test the decompression time too. But the above test is still a good indication. Way too many scripts out there use the default parameters and substitute gzip with gzip2 without any testing. That’s obviously asking for trouble.