On Wednesday 02 February 2005 17:04, Rudi Chiarito wrote: > On Wed, Feb 02, 2005 at 10:32:41AM -0500, Sean Middleditch wrote: > > > > I think staying with gzip is ok as it really is a good middle ground > > > > between speed and disk compression ratio. bzip2 "feels" noticable > > > > slower. > > > > > > Is that true on extraction or only on creation? > > > > Both. > > Then there's LZMA, which is even slower than bzip2 on compression, but > faster and less memory-hungry on decompression. It also tends to > obtain higher compression ratios and has a multithreaded implementation. To show some numbers: this is the current binary kernel LZMA default [ronny@bserv tmp]$ /usr/bin/time ./lzma <kernel.cpio >kernel.cpio.lzma 92.08user 0.31system 1:32.59elapsed 99%CPU (0avgtext+0avgdata 0maxresident)k 0inputs+0outputs (0major+3809minor)pagefaults 0swaps [ronny@bserv tmp]$ ls -lh kernel.cpio.lzma -rw-rw-r-- 1 ronny ronny 12M 2. Feb 21:02 kernel.cpio.lzma [ronny@bserv tmp]$ /usr/bin/time ./lzma -d <kernel.cpio.lzma >/dev/null 5.02user 0.02system 0:05.05elapsed 100%CPU (0avgtext+0avgdata 0maxresident)k 0inputs+0outputs (0major+997minor)pagefaults 0swaps LZMA best compression [ronny@bserv tmp]$ /usr/bin/time ./lzma -x <kernel.cpio >kernel.cpio.lzma 122.63user 0.29system 2:02.95elapsed 99%CPU (0avgtext+0avgdata 0maxresident)k 0inputs+0outputs (0major+3809minor)pagefaults 0swaps [ronny@bserv tmp]$ ls -lh kernel.cpio.lzma -rw-rw-r-- 1 ronny ronny 12M 2. Feb 21:06 kernel.cpio.lzma [ronny@bserv tmp]$ /usr/bin/time ./lzma -d <kernel.cpio.lzma >/dev/null 5.03user 0.01system 0:05.04elapsed 100%CPU (0avgtext+0avgdata 0maxresident)k 0inputs+0outputs (0major+997minor)pagefaults 0swaps GZIP default [ronny@bserv tmp]$ /usr/bin/time gzip <kernel.cpio >kernel.cpio.gz 5.52user 0.22system 0:05.76elapsed 99%CPU (0avgtext+0avgdata 0maxresident)k 0inputs+0outputs (0major+162minor)pagefaults 0swaps [ronny@bserv tmp]$ ls -lh kernel.cpio.gz -rw-rw-r-- 1 ronny ronny 16M 2. Feb 21:07 kernel.cpio.gz [ronny@bserv tmp]$ /usr/bin/time gzip -d <kernel.cpio.gz >/dev/null 0.75user 0.00system 0:00.76elapsed 99%CPU (0avgtext+0avgdata 0maxresident)k 0inputs+0outputs (0major+129minor)pagefaults 0swaps GZIP -9 [ronny@bserv tmp]$ /usr/bin/time gzip -9 <kernel.cpio >kernel.cpio.gz 17.38user 0.21system 0:17.61elapsed 99%CPU (0avgtext+0avgdata 0maxresident)k 0inputs+0outputs (0major+162minor)pagefaults 0swaps [ronny@bserv tmp]$ ls -lh kernel.cpio.gz -rw-rw-r-- 1 ronny ronny 16M 2. Feb 21:08 kernel.cpio.gz [ronny@bserv tmp]$ /usr/bin/time gzip -d <kernel.cpio.gz >/dev/null 0.74user 0.01system 0:00.75elapsed 99%CPU (0avgtext+0avgdata 0maxresident)k 0inputs+0outputs (0major+129minor)pagefaults 0swaps BZIP2 default (-9) [ronny@bserv tmp]$ /usr/bin/time bzip2 <kernel.cpio >kernel.cpio.bz2 16.79user 0.20system 0:17.04elapsed 99%CPU (0avgtext+0avgdata 0maxresident)k 0inputs+0outputs (0major+1741minor)pagefaults 0swaps [ronny@bserv tmp]$ ls -lh kernel.cpio.bz2 -rw-rw-r-- 1 ronny ronny 14M 2. Feb 21:09 kernel.cpio.bz2 [ronny@bserv tmp]$ /usr/bin/time bzip2 -d <kernel.cpio.bz2 >/dev/null 6.20user 0.04system 0:06.25elapsed 99%CPU (0avgtext+0avgdata 0maxresident)k 0inputs+0outputs (0major+1016minor)pagefaults 0swaps BZIP2 -1 [ronny@bserv tmp]$ /usr/bin/time bzip2 -1 <kernel.cpio >kernel.cpio.bz2 13.78user 0.21system 0:14.00elapsed 99%CPU (0avgtext+0avgdata 0maxresident)k 0inputs+0outputs (0major+395minor)pagefaults 0swaps [ronny@bserv tmp]$ ls -lh kernel.cpio.bz2 -rw-rw-r-- 1 ronny ronny 16M 2. Feb 21:12 kernel.cpio.bz2 [ronny@bserv tmp]$ /usr/bin/time bzip2 -d <kernel.cpio.bz2 >/dev/null 4.12user 0.03system 0:04.16elapsed 99%CPU (0avgtext+0avgdata 0maxresident)k 0inputs+0outputs (0major+230minor)pagefaults 0swaps Looking at these numbers I think gzip is still the best compromise -- http://LinuxWiki.org/RonnyBuchmann
