* the difference between btt and blkparse
@ 2010-03-01 11:26 Xuekun Hu
2010-03-01 19:20 ` Alan D. Brunelle
2010-03-02 3:44 ` Xuekun Hu
0 siblings, 2 replies; 3+ messages in thread
From: Xuekun Hu @ 2010-03-01 11:26 UTC (permalink / raw)
To: linux-btrace
Hi, Guys
From the doc, the standard work-flow for btt is to handle the binary
output from "blkparse -d". I found btt can also directly handle the
raw trace file generated by blktrace. However I found the two results
are not match.
For example,
1. "dd if=/dev/urandom of=test" on /dev/sda1
2. blktrace -d /dev/sda -o trace
3. blkparse -i trace.blktrace.* -d bp.bin
4. btt -i bp.bin
Outputs from "btt -i bp.bin"
========== All Devices ==========
ALL MIN AVG MAX N
--------------- ------------- ------------- ------------- -----------
Q2Q 0.000000475 0.000064279 0.365450177 5779
Q2G 0.000000558 0.000001273 0.000020606 1168
G2I 0.000000153 0.000000348 0.000001737 1168
Q2M 0.000000126 0.000000253 0.000018475 91312
I2D 0.000001246 0.032353034 0.174700914 1168
M2D 0.000001785 0.043371005 0.161936556 91312
D2C 0.000176612 0.147189282 0.369407869 5780
Q2C 0.000179449 0.190421403 0.369435695 5780
5. btt -i trace.blktrace.*
outputs from "btt -i trace.blktrace.*"
========== All Devices ==========
ALL MIN AVG MAX N
--------------- ------------- ------------- ------------- -----------
Q2Q 0.000000497 0.000094690 0.366905512 3923
Q2G 0.000000643 0.000001384 0.000020606 57
G2I 0.000000153 0.000000368 0.000001737 57
Q2M 0.000000126 0.000000263 0.000018475 3867
I2D 0.000001246 0.007264465 0.174700914 57
M2D 0.000001785 0.006241516 0.064217079 3867
D2C 0.000176612 0.125216704 0.369407869 3924
Q2C 0.000179449 0.131473364 0.369435695 3924
At the 4) step, the total number of requests are 5780, at the 5) step,
the total number of requests are 3924.
at the 4) step, Q2G is 1168, Q2M is 91312, from my understanding, Q2G
+ Q2M should equal to Q2C. So seems like the 4) result is not correct,
while 5) result is correct.
Is my understanding right, or is it a blkparse bug?
Thx, Xuekun
^ permalink raw reply [flat|nested] 3+ messages in thread
* Re: the difference between btt and blkparse
2010-03-01 11:26 the difference between btt and blkparse Xuekun Hu
@ 2010-03-01 19:20 ` Alan D. Brunelle
2010-03-02 3:44 ` Xuekun Hu
1 sibling, 0 replies; 3+ messages in thread
From: Alan D. Brunelle @ 2010-03-01 19:20 UTC (permalink / raw)
To: linux-btrace
On Mon, 2010-03-01 at 19:26 +0800, Xuekun Hu wrote:
> Hi, Guys
>
> >From the doc, the standard work-flow for btt is to handle the binary
> output from "blkparse -d". I found btt can also directly handle the
> raw trace file generated by blktrace. However I found the two results
> are not match.
'btt' is intended only to be run w/ the output from blkparse -d : the
fact that "it kind of works" with straight blktrace output only means
that the results will be "kind of right". :-)
Regards,
Alan
^ permalink raw reply [flat|nested] 3+ messages in thread
* Re: the difference between btt and blkparse
2010-03-01 11:26 the difference between btt and blkparse Xuekun Hu
2010-03-01 19:20 ` Alan D. Brunelle
@ 2010-03-02 3:44 ` Xuekun Hu
1 sibling, 0 replies; 3+ messages in thread
From: Xuekun Hu @ 2010-03-02 3:44 UTC (permalink / raw)
To: linux-btrace
Oh. I think I misunderstood the option usage.
I orginally used "blkparse -i trace.blkparse.* -d bp.bin". However it
handled number of cpu times, instead of one. "blkparse -i
trace.blkparse.0" and "blkparse -i trace.blkparse.1" have the same
results.
While for btt, I orginally used "btt -i trace.blkparse.*". However the
result was only to handle trace.blkparse.0, instead of others. So the
two results were not match.
Now I will stick to the standard workflow.
Thanks for your help.
On Tue, Mar 2, 2010 at 3:20 AM, Alan D. Brunelle <Alan.Brunelle@hp.com> wrote:
> On Mon, 2010-03-01 at 19:26 +0800, Xuekun Hu wrote:
>> Hi, Guys
>>
>> >From the doc, the standard work-flow for btt is to handle the binary
>> output from "blkparse -d". I found btt can also directly handle the
>> raw trace file generated by blktrace. However I found the two results
>> are not match.
>
> 'btt' is intended only to be run w/ the output from blkparse -d : the
> fact that "it kind of works" with straight blktrace output only means
> that the results will be "kind of right". :-)
>
> Regards,
> Alan
>
>
^ permalink raw reply [flat|nested] 3+ messages in thread
end of thread, other threads:[~2010-03-02 3:44 UTC | newest]
Thread overview: 3+ messages (download: mbox.gz follow: Atom feed
-- links below jump to the message on this page --
2010-03-01 11:26 the difference between btt and blkparse Xuekun Hu
2010-03-01 19:20 ` Alan D. Brunelle
2010-03-02 3:44 ` Xuekun Hu
This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox