Performance issues

Vladimir A. Pavlov pv4 at
Thu Jul 12 10:50:48 PDT 2007

On Thursday 12 July 2007 18:51, Zoran Kikic wrote:
> time tar -xf linux-
> real    10m50.946s
> user    0m29.242s
> sys     0m3.536s

On my AMD Athlon XP 2500+ (1833MHz) I have

time tar -xf ~/arc/linux-2.6.20.tar.bz2 

real    0m51.626s
user    0m31.260s
sys     0m3.310s

The process actually runs "user (the process itself) + sys (system
calls)" seconds. As you can see your values very like my ones. That's
probably because the bottleneck is the hard disk access speed (I have
a 7200 RPM hard drive).

Your "real" time value about 11 minutes may be connected with you
running many other applications eating resources of your system
(processor/disk/memory) (that can be, for example, another unpacking

The simple (but maybe incorrect a little) is following.

Let's assume you run two processes of unpacking different files of
about the same size at the same time. Every one of them requires 30s if
being the only running process in the system. But they run
simultaneously so the kernel tries to distribute process time between

That results in both of the processes being finished about at the same
time. Since every of them requires 30s to unpack a file, both require
60s no matter whether you run them simultaneous or one by one. So the
"real" time of both the processes would be 60s.

But every process runs just a half of the whole time, so (user + sys)
would be 30s for each process and you would have nearly what I have in
my test above (though I actually run only one process of unpacking, at
the same time looking web pages in browser and loading mail with

> When I compiled my kernel the first time (before reboot)
> I did not recognized such problems..

That's because you didn't run many other programs then.

Nothing but perfection


More information about the lfs-support mailing list