Dariush Debian Diary

Contenders:

simfs                 112G   26G   87G  23% /
/dev/shm              2.0G  369M  1.7G  18% /home/eyck/ramdysk

Fight:

export CONCURRENCY_LEVEL=8

eyck@etchdev386:~/40m-ovz/work$ time nice rm -r linux* patches Patching.log && time sh rediff.sh
...
make[1]: Leaving directory `/home/eyck/40m-ovz/work/linux-2.6.18'
====== making target stamp-kernel-headers [new prereqs: linux-headers-2.6.18.bsd40l-vz linux-headers-2.6.18.bsd40l-vz]======
This is kernel package version 10.067.
echo done > stamp-kernel-headers
====== making target kernel_headers [new prereqs: stamp-configure stamp-prepare stamp-kernel-headers]======
This is kernel package version 10.067.

real    17m19.657s
user    28m30.859s
sys     6m11.603s

real    18m3.658s
user    28m44.144s
sys     6m40.757s

export CONCURRENCY_LEVEL=8

eyck@etchdev386:~/ramdysk/40m-ovz/work$ time nice rm -r linux* patches Patching.log && time sh rediff.sh
....
====== making target stamp-kernel-headers [new prereqs: linux-headers-2.6.18.bsd40l-vz linux-headers-2.6.18.bsd40l-vz]======
This is kernel package version 10.067.
echo done > stamp-kernel-headers
====== making target kernel_headers [new prereqs: stamp-configure stamp-prepare stamp-kernel-headers]======
This is kernel package version 10.067.

real    17m19.916s
user    28m14.074s
sys     5m35.077s

real    17m56.725s
user    28m26.523s
sys     5m52.778s

Soo... it looks like (since this machine has 14G ram) it managed too cache all the files, and there's no visible advantage to using ramdisc. And now without openvz overhead:

export CONCURRENCY_LEVEL=8
....
make[1]: Leaving directory `/home/eyck/ramdysk/40m-ovz/work/linux-2.6.18'
====== making target stamp-kernel-headers [new prereqs: linux-headers-2.6.18.bsd40l-vz linux-headers-2.6.18.bsd40l-vz]======
This is kernel package version 10.067.
echo done > stamp-kernel-headers
====== making target kernel_headers [new prereqs: stamp-configure stamp-prepare stamp-kernel-headers]======
This is kernel package version 10.067.

real    17m36.922s
user    28m8.894s
sys     5m30.633s

real    18m13.800s
user    28m21.430s
sys     5m48.630s

export CONCURRENCY_LEVEL=8
....

make[1]: Leaving directory `/home/eyck/40m-ovz/work/linux-2.6.18'
====== making target stamp-kernel-headers [new prereqs: linux-headers-2.6.18.bsd40l-vz linux-headers-2.6.18.bsd40l-vz]======
This is kernel package version 10.067.
echo done > stamp-kernel-headers
====== making target kernel_headers [new prereqs: stamp-configure stamp-prepare stamp-kernel-headers]======
This is kernel package version 10.067.

real    17m55.724s
user    28m26.759s
sys     6m2.363s

real    18m45.132s
user    28m39.971s
sys     6m31.328s

eyck@codev64:~/40m-ovz/work$ export CONCURRENCY_LEVEL=16 eyck@codev64:~/40m-ovz/work$ time nice rm -r linux* patches Patching.log && time sh rediff.sh cp -pf debian/control.dist debian/control make[1]: Leaving directory `/home/eyck/40m-ovz/work/linux-2.6.18' ====== making target stamp-kernel-headers [new prereqs: linux-headers-2.6.18.bsd40l-vz linux-headers-2.6.18.bsd40l-vz]====== This is kernel package version 10.067. echo done > stamp-kernel-headers ====== making target kernel_headers [new prereqs: stamp-configure stamp-prepare stamp-kernel-headers]====== This is kernel package version 10.067. real 20m5.180s user 28m41.280s sys 5m57.190s real 20m52.680s user 28m54.396s sys 6m26.556s eyck@codev64:~/40m-ovz/work$ export CONCURRENCY_LEVEL=4 ... ====== making target stamp-kernel-headers [new prereqs: linux-headers-2.6.18.bsd40l-vz linux-headers-2.6.18.bsd40l-vz]====== This is kernel package version 10.067. echo done > stamp-kernel-headers ====== making target kernel_headers [new prereqs: stamp-configure stamp-prepare stamp-kernel-headers]====== This is kernel package version 10.067. real 15m11.809s user 28m24.991s sys 5m47.362s real 15m58.245s user 28m37.815s sys 6m16.424s eyck@codev64:~/40m-ovz/work$ export CONCURRENCY_LEVEL=2 ... ====== making target kernel_headers [new prereqs: stamp-configure stamp-prepare stamp-kernel-headers]====== This is kernel package version 10.067. real 21m25.527s user 27m58.473s sys 5m22.156s real 22m9.967s user 28m11.438s sys 5m51.326s eyck@etchdev386:~/40m-ovz/work$ export CONCURRENCY_LEVEL=6 ... echo done > stamp-kernel-headers ====== making target kernel_headers [new prereqs: stamp-configure stamp-prepare stamp-kernel-headers]====== This is kernel package version 10.067. real 15m8.466s user 28m36.763s sys 6m10.971s real 15m52.911s user 28m49.976s sys 6m40.661s real 16m3.678s user 28m50.100s sys 6m44.369s eyck@etchdev386:~/40m-ovz/work$ export CONCURRENCY_LEVEL=8 echo done > stamp-kernel-headers ====== making target kernel_headers [new prereqs: stamp-configure stamp-prepare stamp-kernel-headers]====== This is kernel package version 10.067. real 17m56.831s user 28m18.070s sys 6m6.831s real 18m41.792s user 28m31.331s sys 6m36.289s real 18m53.592s user 28m31.447s sys 6m39.985s
[/Ugly notepad] permanent link

This is a rant about 'common knowledge' and about how it hurts technology. You would think that what I'm talking here are wiki's...well.. maybe a bit. Wiki's posess one truly great ?thing? going on for them - they're not authoritary sources and it's very easy to correct them. And most importantly - you're generally encouraged to fix wiki articles. This results in very interesting pattern - often you'll find wiki article that describes something in great detail... and then goes on debunking previous claims by prividing correct definitions PLUS reasons for the not-totally-correct explanation that prevails. There is very well known and recurring pattern out there, small fact, like '2.4.x kernels won't compile with 3.x gcc', they are repeated few times by those that tested this, then the whole 'clergy' learns it by heart ( those that have no idea what they're talking about, but repeat what they heard/read from their 'guru'). By repeating their mantra, they enter 'inner circle' of their religion, they feel superior to the uninitiated, closer to their 'gods' etc... This are all well known ...
[/Ugly notepad] permanent link

how about - squid configured with null cache, using only some memory, and oops attached to it as parent?
[/Ugly notepad] permanent link

#------------------------------------------------

See more ...

[/Ugly notepad] permanent link

Kopia strony by tokarz:

See more ...

[/Ugly notepad] permanent link