Jake Hamby <[EMAIL PROTECTED]> wrote:
> Okay, I did a more "scientific" test and uncompressed the archive first.
> Then I ran every version of tar at least twice (gtar three times b/c the
> first two runs were so different). I ran star at the end to make sure the
> results weren't getting slower over time, and star ran even faster than the
> initial runs, the clear winner by 36-48 seconds. Solaris tar placed second,
> with gnutar and bsdtar slightly (and unpredictably) slower than "tar".
>
> Another interesting thing to note is that star uses 50% more system time than
> Solaris tar, despite finishing faster. Since it is reading the same amount
> of data and creating the same destination files and directories, I'm curious
> why Solaris would take more CPU time to run star and yet finish faster. Time
> for me to practice my dtrace skills.
>
> Test system: 2.8GHz Pentium 4, 512MB DDR333 RAM, 120GB Seagate ST3120026A
> drive, OpenSolaris w/ JDS and 2 terminal windows open, bonnie++ gives read
> speed as 50588K/sec, write speed as 44935K/sec.
>
> --
> Jake Hamby
>
> $ bzcat ~/Downloads/kde/KDEkderequired-341.tar.bz2 >KDEkderequired-341.tar
> $ sync
> $ time star xf KDEkderequired-341.tar
> star: 72176 blocks + 4608 bytes (total of 739086848 bytes = 721764.50k).
>
> real 2m53.636s
> user 0m1.280s
> sys 0m32.205s
> $ rm -r KDEkderequired-341
> $ sync
> $ time star xf KDEkderequired-341.tar
> star: 72176 blocks + 4608 bytes (total of 739086848 bytes = 721764.50k).
star never tried to be the fastest with extract, this is because star
tries to bethe most failsafe tar. Star calls fsync(2) for exvery extracted
file. If you like to compare with other tar implementations, you would need
to run "star -no-fsync ....."
Jörg
--
EMail:[EMAIL PROTECTED] (home) Jörg Schilling D-13353 Berlin
[EMAIL PROTECTED] (uni)
[EMAIL PROTECTED] (work) Blog: http://schily.blogspot.com/
URL: http://cdrecord.berlios.de/old/private/ ftp://ftp.berlios.de/pub/schily
_______________________________________________
opensolaris-discuss mailing list
opensolaris-discuss@opensolaris.org
