Skip to contentSkip to conversation

2.5.61-mm1 +/- as or cfq with contest

BENCHMARK Con Kolivas 3 participants 6 messages Feb 16, 2003

lore
Share & tools
lore
#0 lore raw

Here are contest (http://contest.kolivas.org) results with osdl
(http://www.osdl.org) hardware for 2.5.61-mm1 with either the as i/o scheduler or the cfq scheduler.

no_load:

Kernel         [runs]   Time    CPU%    Loads   LCPU%   Ratio
2.5.60-mm1          3   79      94.9    0.0     0.0     1.00
2.5.61              1   79      94.9    0.0     0.0     1.00
2.5.61-mm1          3   81      91.4    0.0     0.0     1.00
2.5.61-mm1cfq       3   79      94.9    0.0     0.0     1.00
cacherun:
Kernel         [runs]   Time    CPU%    Loads   LCPU%   Ratio
2.5.60-mm1          3   75      100.0   0.0     0.0     0.95
2.5.61              1   76      97.4    0.0     0.0     0.96
2.5.61-mm1          3   76      97.4    0.0     0.0     0.94
2.5.61-mm1cfq       3   76      97.4    0.0     0.0     0.96
process_load:
Kernel         [runs]   Time    CPU%    Loads   LCPU%   Ratio
2.5.60-mm1          3   91      78.0    32.3    18.7    1.15
2.5.61              1   93      80.6    29.0    16.1    1.18
2.5.61-mm1          3   179     41.9    178.0   57.0    2.21
2.5.61-mm1cfq       3   188     39.9    196.7   59.0    2.38
ctar_load:
Kernel         [runs]   Time    CPU%    Loads   LCPU%   Ratio
2.5.60-mm1          3   98      79.6    1.0     4.1     1.24
2.5.61              2   100     79.0    1.0     4.0     1.27
2.5.61-mm1          2   137     58.4    2.0     5.8     1.69
2.5.61-mm1cfq       3   104     76.0    1.0     3.8     1.32
xtar_load:
Kernel         [runs]   Time    CPU%    Loads   LCPU%   Ratio
2.5.60-mm1          3   108     70.4    1.0     3.7     1.37
2.5.61              2   102     75.5    1.0     4.9     1.29
2.5.61-mm1          2   158     48.7    2.0     4.4     1.95
2.5.61-mm1cfq       3   104     74.0    1.0     3.8     1.32
io_load:
Kernel         [runs]   Time    CPU%    Loads   LCPU%   Ratio
2.5.60-mm1          3   112     67.0    15.7    7.1     1.42
2.5.61              2   143     52.4    32.9    13.3    1.81
2.5.61-mm1          2   634     12.5    257.3   24.6    7.83
2.5.61-mm1cfq       3   397     19.6    123.3   18.1    5.03
io_other:
Kernel         [runs]   Time    CPU%    Loads   LCPU%   Ratio
2.5.60-mm1          3   89      84.3    10.5    5.6     1.13
2.5.61              2   91      82.4    11.1    5.5     1.15
2.5.61-mm1          2   187     41.7    84.7    27.3    2.31
2.5.61-mm1cfq       3   199     39.2    77.2    23.5    2.52
read_load:
Kernel         [runs]   Time    CPU%    Loads   LCPU%   Ratio
2.5.60-mm1          3   93      81.7    2.8     2.2     1.18
2.5.61              2   102     77.5    6.3     4.9     1.29
2.5.61-mm1          2   120     65.8    8.9     5.8     1.48
2.5.61-mm1cfq       3   109     72.5    7.1     5.5     1.38
list_load:
Kernel         [runs]   Time    CPU%    Loads   LCPU%   Ratio
2.5.60-mm1          3   96      79.2    0.0     6.2     1.22
2.5.61              2   95      81.1    0.0     6.3     1.20
2.5.61-mm1          2   97      79.4    0.0     6.2     1.20
2.5.61-mm1cfq       3   97      79.4    0.0     6.2     1.23
mem_load:
Kernel         [runs]   Time    CPU%    Loads   LCPU%   Ratio
2.5.60-mm1          3   95      82.1    51.7    2.1     1.20
2.5.61              1   96      81.2    54.0    2.1     1.22
2.5.61-mm1          2   128     61.7    72.0    2.3     1.58
2.5.61-mm1cfq       3   117     66.7    61.0    1.7     1.48
dbench_load:
Kernel         [runs]   Time    CPU%    Loads   LCPU%   Ratio
2.5.61              2   237     32.5    3.0     47.3    3.00
2.5.61-mm1          2   716     10.8    11.0    50.4    8.84
2.5.61-mm1cfq       3   426     18.1    5.7     50.7    5.39

So we're getting into a situation where 2.6 will have either server or desktop tuning? I guess if one can't fit all (ideal) then this is a good compromise. What I don't understand is why the anticipatory scheduler takes longer to compile a kernel without any load running? This happened with previous tests of the as scheduler too.

Con

#1 lore raw
Quoted 10 lines · Con Kolivas

Here are contest (http://contest.kolivas.org) results with osdl
(http://www.osdl.org) hardware for 2.5.61-mm1 with either the as i/o scheduler or the cfq scheduler.

io_load:

Kernel         [runs]   Time    CPU%    Loads   LCPU%   Ratio
2.5.60-mm1          3   112     67.0    15.7    7.1     1.42
2.5.61              2   143     52.4    32.9    13.3    1.81
2.5.61-mm1          2   634     12.5    257.3   24.6    7.83
2.5.61-mm1cfq       3   397     19.6    123.3   18.1    5.03

These loo fishy, could be some other interaction. I'm consistently beating 2.5.60-mm1/2.5.61 on io_load here, but that is 2.5.61 base and not 2.5.61-mm1 base. Could be something odd happening there.

Signature · 1 line
Jens Axboe
#2 lore raw
Quoted 12 lines · Jens Axboe
Quoted 10 lines · Con Kolivas

Here are contest (http://contest.kolivas.org) results with osdl
(http://www.osdl.org) hardware for 2.5.61-mm1 with either the as i/o scheduler or the cfq scheduler.

io_load:

Kernel         [runs]   Time    CPU%    Loads   LCPU%   Ratio
2.5.60-mm1          3   112     67.0    15.7    7.1     1.42
2.5.61              2   143     52.4    32.9    13.3    1.81
2.5.61-mm1          2   634     12.5    257.3   24.6    7.83
2.5.61-mm1cfq       3   397     19.6    123.3   18.1    5.03
Quoted 3 lines

These loo fishy, could be some other interaction. I'm consistently beating 2.5.60-mm1/2.5.61 on io_load here, but that is 2.5.61 base and not 2.5.61-mm1 base. Could be something odd happening there.

I dont think they're fishy - taken in the mm1 context -. I have tested cfq3a without mm1 and it does beat the baseline. See a previous email I posted with it.

Con

#3 lore raw
Quoted 13 lines · Con Kolivas
Quoted 12 lines · Jens Axboe
Quoted 10 lines · Con Kolivas

Here are contest (http://contest.kolivas.org) results with osdl
(http://www.osdl.org) hardware for 2.5.61-mm1 with either the as i/o scheduler or the cfq scheduler.

io_load:

Kernel         [runs]   Time    CPU%    Loads   LCPU%   Ratio
2.5.60-mm1          3   112     67.0    15.7    7.1     1.42
2.5.61              2   143     52.4    32.9    13.3    1.81
2.5.61-mm1          2   634     12.5    257.3   24.6    7.83
2.5.61-mm1cfq       3   397     19.6    123.3   18.1    5.03
Quoted 7 lines
Quoted 3 lines

These loo fishy, could be some other interaction. I'm consistently beating 2.5.60-mm1/2.5.61 on io_load here, but that is 2.5.61 base and not 2.5.61-mm1 base. Could be something odd happening there.

I dont think they're fishy - taken in the mm1 context -. I have tested cfq3a without mm1 and it does beat the baseline. See a previous email I posted with it.

I didn't mean that you have done something fishy, but that there's a fishy interaction between -mm + CFQ :)

Signature · 1 line
Jens Axboe
#4 lore raw
Quoted 19 lines · Jens Axboe
Quoted 17 lines · Con Kolivas
Quoted 15 lines · Jens Axboe
Quoted 10 lines · Con Kolivas

Here are contest (http://contest.kolivas.org) results with osdl
(http://www.osdl.org) hardware for 2.5.61-mm1 with either the as i/o scheduler or the cfq scheduler.

io_load:

Kernel         [runs]   Time    CPU%    Loads   LCPU%   Ratio
2.5.60-mm1          3   112     67.0    15.7    7.1     1.42
2.5.61              2   143     52.4    32.9    13.3    1.81
2.5.61-mm1          2   634     12.5    257.3   24.6    7.83
2.5.61-mm1cfq       3   397     19.6    123.3   18.1    5.03

These loo fishy, could be some other interaction. I'm consistently beating 2.5.60-mm1/2.5.61 on io_load here, but that is 2.5.61 base and not 2.5.61-mm1 base. Could be something odd happening there.

Quoted 7 lines
Quoted 3 lines

I dont think they're fishy - taken in the mm1 context -. I have tested cfq3a without mm1 and it does beat the baseline. See a previous email I posted with it.

I didn't mean that you have done something fishy, but that there's a fishy interaction between -mm + CFQ :)

It is the CPU scheduler patch. Con has eariler shown that this patch shoots io_load in the head. 2.5.60-mm1 did not have that patch.

#5 lore raw
Quoted 24 lines · Andrew Morton
Quoted 23 lines · Jens Axboe
Quoted 20 lines · Con Kolivas
Quoted 12 lines · Jens Axboe
Quoted 10 lines · Con Kolivas

Here are contest (http://contest.kolivas.org) results with osdl
(http://www.osdl.org) hardware for 2.5.61-mm1 with either the as i/o scheduler or the cfq scheduler.

io_load:

Kernel         [runs]   Time    CPU%    Loads   LCPU%   Ratio
2.5.60-mm1          3   112     67.0    15.7    7.1     1.42
2.5.61              2   143     52.4    32.9    13.3    1.81
2.5.61-mm1          2   634     12.5    257.3   24.6    7.83
2.5.61-mm1cfq       3   397     19.6    123.3   18.1    5.03
Quoted 3 lines

These loo fishy, could be some other interaction. I'm consistently beating 2.5.60-mm1/2.5.61 on io_load here, but that is 2.5.61 base and not 2.5.61-mm1 base. Could be something odd happening there.

I dont think they're fishy - taken in the mm1 context -. I have tested cfq3a without mm1 and it does beat the baseline. See a previous email I posted with it.

Quoted 6 lines
Quoted 3 lines

I didn't mean that you have done something fishy, but that there's a fishy interaction between -mm + CFQ :)

It is the CPU scheduler patch. Con has eariler shown that this patch shoots io_load in the head. 2.5.60-mm1 did not have that patch.

and process_load, and dbench_load :)

Thanks, makes sense.

Signature · 1 line
Jens Axboe