query: [PATCH 2/2] cgroup: Remove call to synchronize_rcu incgroup_attach_task

[Mike Galbraith]

Greetings,

Wrt these patches:

https://lkml.org/lkml/2010/11/24/14 [PATCH 1/2] cgroup: Set CGRP_RELEASABLE when adding to a cgroup
https://lkml.org/lkml/2010/11/24/15 [PATCH 2/2] cgroup: Remove call to synchronize_rcu in cgroup_attach_task

I received a query regarding 2/2 because a large database company is
apparently moving tasks between cgroups frequently enough that their
database initialization time dropped from ~11 hours to ~4 hours when
they applied this patch.

Curious why these got no traction.

Regardless of the wisdom of frequent cgroup moves, Joe User launching
light tasks via cgexec would gain quite a bit as well it seems.

Q/D measurement of cgexec -g cpu:test sh -c exit.

Unpatched:

marge:~ # time perf stat --repeat=1000 cgexec -g cpu:test sh -c exit

 Performance counter stats for 'cgexec -g cpu:test sh -c exit' (1000 runs):

          2.686149 task-clock-msecs         #      0.119 CPUs    ( +-   0.067% )
                 2 context-switches         #      0.001 M/sec   ( +-   0.100% )
                 0 CPU-migrations           #      0.000 M/sec   ( +-  35.231% )
               633 page-faults              #      0.236 M/sec   ( +-   0.003% )
           6292755 cycles                   #   2342.668 M/sec   ( +-   0.062% )
           4844702 instructions             #      0.770 IPC     ( +-   0.004% )
            974011 branches                 #    362.605 M/sec   ( +-   0.005% )
             35187 branch-misses            #      3.613 %       ( +-   0.027% )
     <not counted> cache-references        
     <not counted> cache-misses            

        0.022597372  seconds time elapsed   ( +-   0.981% )


real    0m23.054s
user    0m0.380s
sys     0m0.000s

Patched:

marge:~ # time perf stat --repeat=1000 cgexec -g cpu:test sh -c exit

 Performance counter stats for 'cgexec -g cpu:test sh -c exit' (1000 runs):

          2.639303 task-clock-msecs         #      0.915 CPUs    ( +-   0.051% )
                 1 context-switches         #      0.000 M/sec   ( +-   1.048% )
                 0 CPU-migrations           #      0.000 M/sec   ( +-  21.602% )
               633 page-faults              #      0.240 M/sec   ( +-   0.003% )
           3883572 cycles                   #   1471.439 M/sec   ( +-   1.767% )
           4830809 instructions             #      1.244 IPC     ( +-   0.191% )
            975204 branches                 #    369.493 M/sec   ( +-   0.009% )
             35314 branch-misses            #      3.621 %       ( +-   0.060% )
     <not counted> cache-references        
     <not counted> cache-misses            

        0.002884424  seconds time elapsed   ( +-   0.055% )


real    0m3.329s
user    0m0.012s
sys     0m0.352s



--
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@xxxxxxxxxxxxxxx
More majordomo info at  http://vger.kernel.org/majordomo-info.html
Please read the FAQ at  http://www.tux.org/lkml/