[Date Prev][Date Next]   [Thread Prev][Thread Next]   [Thread Index] [Date Index] [Author Index]

Re: rpm slowness



Jeff Johnson <jbj@redhat.com> writes:

> > Besides that, is anyone looking for improvements in the rpm performance?
> > Because my feeling is that it's getting more slower (I mean really) after
> > each release...
> > 
> > If someone has any benchmarks to send, it would be interesting to take a
> > look at them (maybe comparing rpm-3, rpm-4{0,1}, dpkg...)
> 
> I compile rpm with -pg every 3 months or so, and I know where
> all the bottlenecks are.
> 
> I know most of the fixes too. :-)
> 
> FWIW, rpm is not getting slower. Go try rpm-4.0.2 and see what you think.

The core engine may not be getting slower, but default policies are
making rpm -seem- slower for the average user.  The most common
example is slowness of 'rpm -qa' in RPM 4.1 and higher.

[cturner@min cturner]$ time /usr/bin/rpmquery -qa > /dev/null

real    0m17.062s
user    0m16.629s
sys     0m0.275s
[cturner@min cturner]$ time /usr/bin/rpmquery -qa --nodigest --nosignature > /dev/null

real    0m0.919s
user    0m0.656s
sys     0m0.256s

Though RPM's core itself and its interactions with berkeley db may not
be slower, from the outside, it is considerably slower to the average
command line user because of changes in default policies by various
distros.

FWIW, you can put --nodigest and --nosignature by default on your 'rpm
-qa' with this ~/.popt macro.  Note it is slightly less reliable
because it disables signature and digest checking, but for all
practical uses, it isn't a huge loss.  Remove --nodigest for about a
2x penalty but with much higher reliability guarantees.

[cturner@min cturner]$ cat ~/.popt
rpm     exec --qa rpmq -qa --nodigest --nosignature
[cturner@min cturner]$

Chip

-- 
Chip Turner                   cturner@redhat.com
                              Red Hat, Inc.





[Date Prev][Date Next]   [Thread Prev][Thread Next]   [Thread Index] [Date Index] [Author Index] []