linux-kernel.vger.kernel.org archive mirror
 help / color / mirror / Atom feed
* OSDL DBT-2 AS vs. Deadline 2.6.0-test4-mm3
@ 2003-09-05 18:27 markw
  2003-09-05 20:17 ` Mike Fedyk
  0 siblings, 1 reply; 3+ messages in thread
From: markw @ 2003-09-05 18:27 UTC (permalink / raw)
  To: linux-kernel

With recent firmware upgrades to our controller cards in our STP system,
I was finally able to make use of Mark Haverkamp's aacraid patch to
avoid bounce buffers.  I've broken out the data by host, since we see
varying results between our system:

AS Scheduler:
host      avg. notpm  stddev  samples
--------  ----------  ------  -------
stp4-000     1273.13   66.91        6  
stp4-001     1210.33   27.81        6
stp4-002     1217.33   10.03        6
stp4-003     1338.67   12.88        5

Deadline Scheduler:
host      avg. notpm  stddev  samples
--------  ----------  ------  -------
stp4-000     1268.50   61.81        3
stp4-001     1232.67   11.15        3
stp4-002     1214.00       -        1
stp4-003     1348.00   30.04        3


Here's a profile comparison:

METRICS OVER LAST 20 MINUTES:
--------------- -------- ----- ---- -------- -----------------------------------
Kernel          Elevator NOTPM CPU% Blocks/s URL                                
--------------- -------- ----- ---- -------- -----------------------------------
2.6.0-test4-mm3 as        1315 90.8   9983.1 http://khack.osdl.org/stp/278809/  
2.6.0-test4-mm3 deadline  1314 90.3  10036.0 http://khack.osdl.org/stp/278903/  

FUNCTIONS SORTED BY LOAD:
-- ------------------------- ------- ------------------------- -------
 # as 2.6.0-test4-mm3        ticks   deadline 2.6.0-test4-mm3  ticks  
-- ------------------------- ------- ------------------------- -------
 1 default_idle              5746007 default_idle              5764451
 2 schedule                    53609 schedule                    52174
 3 scsi_request_fn             29340 do_softirq                  25905
 4 __make_request              25949 scsi_request_fn             23549
 5 do_softirq                  25560 __make_request              20325
 6 scsi_end_request            16524 try_to_wake_up              11723
 7 try_to_wake_up              11939 scsi_end_request            11340
 8 dio_bio_end_io              11253 dio_bio_end_io              10950
 9 do_anonymous_page            8234 do_anonymous_page            8110
10 ipc_lock                     7366 ipc_lock                     6706
11 sys_semtimedop               5493 sys_semtimedop               5152
12 sysenter_past_esp            5080 sysenter_past_esp            5064
13 direct_io_worker             5065 direct_io_worker             4732
14 dio_await_one                4084 dio_await_one                3779
15 __copy_to_user_ll            3643 __mod_timer                  3571
16 __mod_timer                  3582 blk_run_queue                3521
17 blk_run_queue                3411 __copy_to_user_ll            3495
18 __might_sleep                3316 __might_sleep                3277
19 kmem_cache_alloc             3225 kmem_cache_alloc             3054
20 semctl_main                  3072 semctl_main                  3016

-- 
Mark Wong - - markw@osdl.org
Open Source Development Lab Inc - A non-profit corporation
12725 SW Millikan Way - Suite 400 - Beaverton, OR 97005
(503) 626-2455 x 32 (office)
(503) 626-2436      (fax)

^ permalink raw reply	[flat|nested] 3+ messages in thread

* Re: OSDL DBT-2 AS vs. Deadline 2.6.0-test4-mm3
  2003-09-05 18:27 OSDL DBT-2 AS vs. Deadline 2.6.0-test4-mm3 markw
@ 2003-09-05 20:17 ` Mike Fedyk
  2003-09-08  7:35   ` Jens Axboe
  0 siblings, 1 reply; 3+ messages in thread
From: Mike Fedyk @ 2003-09-05 20:17 UTC (permalink / raw)
  To: markw; +Cc: linux-kernel

On Fri, Sep 05, 2003 at 11:27:56AM -0700, markw@osdl.org wrote:
> FUNCTIONS SORTED BY LOAD:
> -- ------------------------- ------- ------------------------- -------
>  # as 2.6.0-test4-mm3        ticks   deadline 2.6.0-test4-mm3  ticks  
> -- ------------------------- ------- ------------------------- -------
>  1 default_idle              5746007 default_idle              5764451
>  2 schedule                    53609 schedule                    52174
>  3 scsi_request_fn             29340 
>                                      do_softirq                  25905
>  4 __make_request              25949 scsi_request_fn             23549
>  5 do_softirq                  25560 __make_request              20325
>  6 scsi_end_request            16524 try_to_wake_up              11723
>  7 try_to_wake_up              11939 scsi_end_request            11340
>  8 dio_bio_end_io              11253 dio_bio_end_io              10950
>  9 do_anonymous_page            8234 do_anonymous_page            8110
> 10 ipc_lock                     7366 ipc_lock                     6706
> 11 sys_semtimedop               5493 sys_semtimedop               5152
> 12 sysenter_past_esp            5080 sysenter_past_esp            5064
> 13 direct_io_worker             5065 direct_io_worker             4732
> 14 dio_await_one                4084 dio_await_one                3779
> 15 __copy_to_user_ll            3643 __mod_timer                  3571
> 16 __mod_timer                  3582 blk_run_queue                3521
> 17 blk_run_queue                3411 __copy_to_user_ll            3495
> 18 __might_sleep                3316 __might_sleep                3277
> 19 kmem_cache_alloc             3225 kmem_cache_alloc             3054
> 20 semctl_main                  3072 semctl_main                  3016
> 

Looking at this quickly, scsi_request_fn, __make_request, and
scsi_end_request are show the most additional cost on AS.  As well as
__copy_to_user_ll and ipc_lock.

^ permalink raw reply	[flat|nested] 3+ messages in thread

* Re: OSDL DBT-2 AS vs. Deadline 2.6.0-test4-mm3
  2003-09-05 20:17 ` Mike Fedyk
@ 2003-09-08  7:35   ` Jens Axboe
  0 siblings, 0 replies; 3+ messages in thread
From: Jens Axboe @ 2003-09-08  7:35 UTC (permalink / raw)
  To: mfedyk; +Cc: markw, linux-kernel

On Fri, Sep 05 2003, Mike Fedyk wrote:
> On Fri, Sep 05, 2003 at 11:27:56AM -0700, markw@osdl.org wrote:
> > FUNCTIONS SORTED BY LOAD:
> > -- ------------------------- ------- ------------------------- -------
> >  # as 2.6.0-test4-mm3        ticks   deadline 2.6.0-test4-mm3  ticks  
> > -- ------------------------- ------- ------------------------- -------
> >  1 default_idle              5746007 default_idle              5764451
> >  2 schedule                    53609 schedule                    52174
> >  3 scsi_request_fn             29340 
> >                                      do_softirq                  25905
> >  4 __make_request              25949 scsi_request_fn             23549
> >  5 do_softirq                  25560 __make_request              20325
> >  6 scsi_end_request            16524 try_to_wake_up              11723
> >  7 try_to_wake_up              11939 scsi_end_request            11340
> >  8 dio_bio_end_io              11253 dio_bio_end_io              10950
> >  9 do_anonymous_page            8234 do_anonymous_page            8110
> > 10 ipc_lock                     7366 ipc_lock                     6706
> > 11 sys_semtimedop               5493 sys_semtimedop               5152
> > 12 sysenter_past_esp            5080 sysenter_past_esp            5064
> > 13 direct_io_worker             5065 direct_io_worker             4732
> > 14 dio_await_one                4084 dio_await_one                3779
> > 15 __copy_to_user_ll            3643 __mod_timer                  3571
> > 16 __mod_timer                  3582 blk_run_queue                3521
> > 17 blk_run_queue                3411 __copy_to_user_ll            3495
> > 18 __might_sleep                3316 __might_sleep                3277
> > 19 kmem_cache_alloc             3225 kmem_cache_alloc             3054
> > 20 semctl_main                  3072 semctl_main                  3016
> > 
> 
> Looking at this quickly, scsi_request_fn, __make_request, and
> scsi_end_request are show the most additional cost on AS.  As well as
> __copy_to_user_ll and ipc_lock.

I think you forgot to include the interesting part of your email.

-- 
Jens Axboe


^ permalink raw reply	[flat|nested] 3+ messages in thread

end of thread, other threads:[~2003-09-08  7:35 UTC | newest]

Thread overview: 3+ messages (download: mbox.gz / follow: Atom feed)
-- links below jump to the message on this page --
2003-09-05 18:27 OSDL DBT-2 AS vs. Deadline 2.6.0-test4-mm3 markw
2003-09-05 20:17 ` Mike Fedyk
2003-09-08  7:35   ` Jens Axboe

This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox;
as well as URLs for NNTP newsgroup(s).