* Re: from raid456 patch, about the tiotest tool
2008-09-02 8:02 ` from raid456 patch, about the tiotest tool rae l
@ 2008-09-03 10:21 ` rae l
0 siblings, 0 replies; 2+ messages in thread
From: rae l @ 2008-09-03 10:21 UTC (permalink / raw)
To: Dan Williams, jmm, mhydra, miku
Cc: linux-raid, James W. Laferriere, Neil Brown, linux-kernel
On Tue, Sep 2, 2008 at 4:02 PM, rae l <crquan@gmail.com> wrote:
> On Wed, Aug 27, 2008 at 2:03 AM, rae l <crquan@gmail.com> wrote:
>> Hello, Dan:
>> I've known your md(raid456) patch just before v2.6.26,
>>
>> http://git.kernel.org/?p=linux/kernel/git/torvalds/linux-2.6.git;a=commitdiff;h=8b3e6cdc53b7f29f7026955d6cb6902a49322a15
>>
>> your md patch referred to tiobench, which is the multi-thread
>> benchmark tool, I'm interested on this tool and I found I ran it
>> differently with your results, so I want to know more about it,
>>
>> from its home page, it only released v0.3.3 on Mar 2002,
>>
>> http://sourceforge.net/projects/tiobench/
>>
>> but from its cvs repository, it has released v0.4.1,
>>
>> http://tiobench.cvs.sourceforge.net/tiobench/tiobench/
>>
>> Could you tell me what version of tiobench did you use? I want to see
>> if that's my configuration fault or I have used wrong program version.
>>
>> And when using v0.4.1 of tiobench from its cvs repository, I found its
>> multiple threads random IO statistics code reported wrong results, and
>> have resolved it by this patch,
>>
>> http://sourceforge.net/tracker/index.php?func=detail&aid=2047536&group_id=3857&atid=303857
>>
>> But its upstream seems have not noticed it yet, so I want to if
>> someone on this mailling-list know tiobench upstream and be able to
>> take this message?
>
> Hope someone can wake up the tiobench developer?
>
Regretfully, the tiotest v0.4.1 still have bugs that I haven't
resolved, even applied my previous patch,
such as this reported on a single SATA disk, simple dd benchmark
report its bandwidth is 70MB/s, do you believe it has 1489.8 MB/s
random read performance?
But it's weird when I'm using systemtap on it,
# stap -ve 'probe kernel.function("vfs_read@fs/read_write.c") { ... }'
it reports all pread successfully return fast,
# ./tiotest -R -d /dev/sdb5 -f 1 -r 1024000 -t 1
Tiotest results for 1 concurrent io threads:
,----------------------------------------------------------------------.
| Item | Time | Rate | Usr CPU | Sys CPU |
+-----------------------+----------+--------------+----------+---------+
| Write 1 MBs | 0.1 s | 8.814 MB/s | 0.0 % | 8.8 % |
| Random Write 4000 MBs | 5.1 s | 785.431 MB/s | 11.2 % | 86.8 % |
| Read 1 MBs | 0.0 s | 1037.344 MB/s | 0.0 % | 0.0 % |
| Random Read 4000 MBs | 2.7 s | 1489.866 MB/s | 27.6 % | 72.2 % |
`----------------------------------------------------------------------'
Tiotest latency results:
,-------------------------------------------------------------------------.
| Item | Average latency | Maximum latency | % >2 sec | % >10 sec |
+--------------+-----------------+-----------------+----------+-----------+
| Write | 0.007 ms | 0.028 ms | 0.00000 | 0.00000 |
| Random Write | 0.004 ms | 0.229 ms | 0.00000 | 0.00000 |
| Read | 0.003 ms | 0.058 ms | 0.00000 | 0.00000 |
| Random Read | 0.002 ms | 0.222 ms | 0.00000 | 0.00000 |
|--------------+-----------------+-----------------+----------+-----------|
| Total | 0.003 ms | 0.229 ms | 0.00000 | 0.00000 |
`--------------+-----------------+-----------------+----------+-----------'
--
Denis Cheng
Linux Application Developer
"One of my most productive days was throwing away 1000 lines of code."
- Ken Thompson.
^ permalink raw reply [flat|nested] 2+ messages in thread