From chemical.aniket at gmail.com  Wed Jul 12 01:33:21 2006
From: chemical.aniket at gmail.com (Aniket Kulkarni)
Date: Wed, 12 Jul 2006 11:03:21 +0530
Subject: [Disksim-users] how to extract trace from my own computer
Message-ID: <c6fab9cb0607112233r3ed8ee83m87c130f9ce80c701@mail.gmail.com>

I have problem to extract trace from my own computer.
Are there special programs to extract it.
Please help me out.
aniket
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.andrew.cmu.edu/pipermail/disksim-users/attachments/20060712/ddb24917/attachment.html>

From xyao at cse.unl.edu  Wed Jul 12 02:55:04 2006
From: xyao at cse.unl.edu (Xiaoyu Yao)
Date: Wed, 12 Jul 2006 01:55:04 -0500
Subject: [Disksim-users] how to extract trace from my own computer
In-Reply-To: <c6fab9cb0607112233r3ed8ee83m87c130f9ce80c701@mail.gmail.com>
References: <c6fab9cb0607112233r3ed8ee83m87c130f9ce80c701@mail.gmail.com>
Message-ID: <44B49CC8.5040606@cse.unl.edu>

Which format of trace do you use?
Please include more details if you want help.

Xiaoyu

Aniket Kulkarni wrote:
> I have problem to extract trace from my own computer.
> Are there special programs to extract it.
> Please help me out.
> aniket
> ------------------------------------------------------------------------
>
> _______________________________________________
> Disksim-users mailing list
> Disksim-users at ece.cmu.edu
> https://sos.ece.cmu.edu/mailman/listinfo/disksim-users
>   



From xyao at cse.unl.edu  Wed Jul 12 04:07:43 2006
From: xyao at cse.unl.edu (Xiaoyu Yao)
Date: Wed, 12 Jul 2006 03:07:43 -0500
Subject: [Disksim-users] how to extract trace from my own computer
In-Reply-To: <c6fab9cb0607120102n1cfa0690yc99140f7ed1f81fa@mail.gmail.com>
References: <c6fab9cb0607112233r3ed8ee83m87c130f9ce80c701@mail.gmail.com>	
	<44B49CC8.5040606@cse.unl.edu>
	<c6fab9cb0607120102n1cfa0690yc99140f7ed1f81fa@mail.gmail.com>
Message-ID: <44B4ADCF.3020300@cse.unl.edu>

Do you want to collect trace yourself from your system?
If that is the case, there are many linux kernel patches can do this 
such as LL_TRACE for 2.4 kernel.

Xiaoyu

Aniket Kulkarni wrote:
> We have trace files in the valid directory like  barracuda.trace,  
> cheetah4LP.trace, cheetah9LP.trace & etc.
> These files having some entries such as request flag, block  
> no.,request size, etc.
> How to extract this information from the hard disk?
> aniket
>
> On 7/12/06, *Xiaoyu Yao* <xyao at cse.unl.edu <mailto:xyao at cse.unl.edu>> 
> wrote:
>
>     Which format of trace do you use?
>     Please include more details if you want help.
>
>     Xiaoyu
>
>     Aniket Kulkarni wrote:
>     > I have problem to extract trace from my own computer.
>     > Are there special programs to extract it.
>     > Please help me out.
>     > aniket
>     >
>     ------------------------------------------------------------------------
>     >
>     > _______________________________________________
>     > Disksim-users mailing list
>     > Disksim-users at ece.cmu.edu <mailto:Disksim-users at ece.cmu.edu>
>     > https://sos.ece.cmu.edu/mailman/listinfo/disksim-users
>     >
>
>



From xyao at cse.unl.edu  Wed Jul 12 14:41:42 2006
From: xyao at cse.unl.edu (Xiaoyu Yao)
Date: Wed, 12 Jul 2006 13:41:42 -0500
Subject: [Disksim-users] how to extract trace from my own computer
In-Reply-To: <c6fab9cb0607120117m3555e9efmdf5e580be246930e@mail.gmail.com>
References: <c6fab9cb0607112233r3ed8ee83m87c130f9ce80c701@mail.gmail.com>	
	<44B49CC8.5040606@cse.unl.edu>	
	<c6fab9cb0607120102n1cfa0690yc99140f7ed1f81fa@mail.gmail.com>	
	<44B4ADCF.3020300@cse.unl.edu>
	<c6fab9cb0607120117m3555e9efmdf5e580be246930e@mail.gmail.com>
Message-ID: <44B54266.3060100@cse.unl.edu>

Hi, Aniket,

I haven't use the validation trace before. It seems that the manual does 
not include detailed information about its format.
But you can check the source code and find the format of the validation 
trace. Based on  my understanding, the one validate trace record is 
composed of the following segments:

* Read/Write Flag
* Buffer Action (Hit/Miss)
* Request Address (blkno)
* Request Size (bcount)
* Service time (Real)
* Next Interarrival time (Real)

Check the following code (from disksim 2.0) for details. Disksim 3.0 
does not change the format of validation trace. Hope this will be helpful.

ioreq_event * iotrace_validate_get_ioreq_event (FILE *tracefile, 
ioreq_event *new)
{
   char line[201];
   char rw;
   double servtime;

   if (fgets(line, 200, tracefile) == NULL) {
      addtoextraq((event *) new);
      return(NULL);
   }
   new->time = simtime + (validate_nextinter / (double) 1000);
   if (sscanf(line, "%c %s %d %d %lf %lf\n", &rw, validate_buffaction, 
&new->blkno, &new->bcount, &servtime, &validate_nextinter) != 6) {
      fprintf(stderr, "Wrong number of arguments for I/O trace event 
type\n");
      exit(0);
   }
   validate_lastserv = servtime / (double) 1000;
   if (rw == 'R') {
      new->flags = READ;
   } else if (rw == 'W') {
      new->flags = WRITE;
   } else {
      fprintf(stderr, "Invalid R/W value: %c\n", rw);
      exit(0);
   }
   new->devno = 0;
   new->buf = 0;
   new->opid = 0;
   new->cause = 0;
   new->busno = 0;
   new->tempint2 = 0;
   new->tempint1 = 0;
   validate_lastblkno = new->blkno;
   validate_lastbcount = new->bcount;
   validate_lastread = new->flags & READ;
   return(new);
}

Aniket Kulkarni wrote:
> please tell me the contents of trace file whose type is validate.
> For ascii trace file following are the contents
> 1.request arrival time
> 2.device number
> 3.block number
> 4.request size
> 5.request flags
> Like that please tell contents of validate trace files(such as 
> barracuda.trace, cheetah9LP.trace)
> aniket
>
>
> On 7/12/06, *Xiaoyu Yao* <xyao at cse.unl.edu <mailto:xyao at cse.unl.edu>> 
> wrote:
>
>     Do you want to collect trace yourself from your system?
>     If that is the case, there are many linux kernel patches can do this
>     such as LL_TRACE for 2.4 kernel.
>
>     Xiaoyu
>
>     Aniket Kulkarni wrote:
>     > We have trace files in the valid directory like   barracuda.trace,
>     > cheetah4LP.trace, cheetah9LP.trace & etc.
>     > These files having some entries such as request flag, block
>     > no.,request size, etc.
>     > How to extract this information from the hard disk?
>     > aniket
>     >
>     > On 7/12/06, *Xiaoyu Yao* <xyao at cse.unl.edu
>     <mailto:xyao at cse.unl.edu> <mailto:xyao at cse.unl.edu
>     <mailto:xyao at cse.unl.edu>>>
>     > wrote:
>     >
>     >     Which format of trace do you use?
>     >     Please include more details if you want help.
>     >
>     >     Xiaoyu
>     >
>     >     Aniket Kulkarni wrote:
>     >     > I have problem to extract trace from my own computer.
>     >     > Are there special programs to extract it.
>     >     > Please help me out.
>     >     > aniket
>     >     >
>     >    
>     ------------------------------------------------------------------------
>     >     >
>     >     > _______________________________________________
>     >     > Disksim-users mailing list
>     >     > Disksim-users at ece.cmu.edu
>     <mailto:Disksim-users at ece.cmu.edu>
>     <mailto:Disksim-users at ece.cmu.edu <mailto:Disksim-users at ece.cmu.edu>>
>     >     > https://sos.ece.cmu.edu/mailman/listinfo/disksim-users
>     >     >
>     >
>     >
>
>



From shahrukh at cs.virginia.edu  Thu Jul 13 15:41:51 2006
From: shahrukh at cs.virginia.edu (Shahrukh Rohinton Tarapore)
Date: Thu, 13 Jul 2006 15:41:51 -0400
Subject: [Disksim-users] subdivide latency
Message-ID: <44B6A1FF.9050500@cs.virginia.edu>

Is it possible to subdivide the latency of a request? For example, if a 
request takes 3ms. then is there a mechanism for knowing that 1.5ms 
where spent on the disk mechanics, 0.5ms. were in some adapter or bus, 
and 1.0ms. some other overhead.  I really just want to be able to 
subdivide the latency between disk mechanics and the datas transmission 
through the logical/physical organization. Thanks.

-- shahrukh


From bucy at ece.cmu.edu  Thu Jul 13 15:50:48 2006
From: bucy at ece.cmu.edu (John Bucy)
Date: Thu, 13 Jul 2006 12:50:48 -0700
Subject: [Disksim-users] subdivide latency
In-Reply-To: <44B6A1FF.9050500@cs.virginia.edu>
References: <44B6A1FF.9050500@cs.virginia.edu>
Message-ID: <4fc1e0430607131250h4a061124g17a0d14f67b505ce@mail.gmail.com>

On 7/13/06, Shahrukh Rohinton Tarapore <shahrukh at cs.virginia.edu> wrote:
> Is it possible to subdivide the latency of a request? For example, if a
> request takes 3ms. then is there a mechanism for knowing that 1.5ms
> where spent on the disk mechanics, 0.5ms. were in some adapter or bus,
> and 1.0ms. some other overhead.  I really just want to be able to
> subdivide the latency between disk mechanics and the datas transmission
> through the logical/physical organization. Thanks.

I don't think that this functionality exists presently though the
usual quick/hacky solution would be to put some prints in.  Also,
remember that a lot of these activities are overlapped with each other
which makes this a little harder to think about.

At a lower level, the diskmodel API does expose this (look for 'struct
dm_mech_acctimes' in diskmodel/dm.h) but that won't give you
cache/controller/bus/etc, just mechanics.



john


From shahrukh at cs.virginia.edu  Thu Jul 13 15:54:43 2006
From: shahrukh at cs.virginia.edu (Shahrukh Rohinton Tarapore)
Date: Thu, 13 Jul 2006 15:54:43 -0400
Subject: [Disksim-users] subdivide latency
In-Reply-To: <4fc1e0430607131250h4a061124g17a0d14f67b505ce@mail.gmail.com>
References: <44B6A1FF.9050500@cs.virginia.edu>
	<4fc1e0430607131250h4a061124g17a0d14f67b505ce@mail.gmail.com>
Message-ID: <44B6A503.4010807@cs.virginia.edu>

Yeah, I was worried i'd lose the controller/cache numbers.  Is it 
possible to turn off all the latencies of the organization on the fly? 
So that I can simulate disks using the current setup of disksim, but 
only see latencies from communication when I want to.  Thanks.

-- shahrukh

John Bucy wrote:

> On 7/13/06, Shahrukh Rohinton Tarapore <shahrukh at cs.virginia.edu> wrote:
>
>> Is it possible to subdivide the latency of a request? For example, if a
>> request takes 3ms. then is there a mechanism for knowing that 1.5ms
>> where spent on the disk mechanics, 0.5ms. were in some adapter or bus,
>> and 1.0ms. some other overhead.  I really just want to be able to
>> subdivide the latency between disk mechanics and the datas transmission
>> through the logical/physical organization. Thanks.
>
>
> I don't think that this functionality exists presently though the
> usual quick/hacky solution would be to put some prints in.  Also,
> remember that a lot of these activities are overlapped with each other
> which makes this a little harder to think about.
>
> At a lower level, the diskmodel API does expose this (look for 'struct
> dm_mech_acctimes' in diskmodel/dm.h) but that won't give you
> cache/controller/bus/etc, just mechanics.
>
>
>
> john



From bucy at ece.cmu.edu  Thu Jul 13 17:20:14 2006
From: bucy at ece.cmu.edu (John Bucy)
Date: Thu, 13 Jul 2006 14:20:14 -0700
Subject: Fwd: [Disksim-users] subdivide latency
In-Reply-To: <4fc1e0430607131413h172c32d7ra0020d73aed7688b@mail.gmail.com>
References: <44B6A1FF.9050500@cs.virginia.edu>
	<4fc1e0430607131250h4a061124g17a0d14f67b505ce@mail.gmail.com>
	<44B6A503.4010807@cs.virginia.edu>
	<4fc1e0430607131413h172c32d7ra0020d73aed7688b@mail.gmail.com>
Message-ID: <4fc1e0430607131420l6e548d5fx7cddef71841159ac@mail.gmail.com>

On 7/13/06, Shahrukh Rohinton Tarapore <shahrukh at cs.virginia.edu> wrote:
> Yeah, I was worried i'd lose the controller/cache numbers.  Is it
> possible to turn off all the latencies of the organization on the fly?
> So that I can simulate disks using the current setup of disksim, but
> only see latencies from communication when I want to.  Thanks.

You could add a global variable somewhere that you condition all your
prints on and then you could turn them on and off on the fly...



john


From bucy at ece.cmu.edu  Mon Jul 17 16:06:17 2006
From: bucy at ece.cmu.edu (John Bucy)
Date: Mon, 17 Jul 2006 13:06:17 -0700
Subject: [Disksim-users] Fwd: [Disksim] Portability patch for disksim 3.0
In-Reply-To: <4fc1e0430607171305g2bdb72c0o377fd5f18cd16fbe@mail.gmail.com>
References: <44B7A5A8.2050003@cs.uit.no>
	<4fc1e0430607171305g2bdb72c0o377fd5f18cd16fbe@mail.gmail.com>
Message-ID: <4fc1e0430607171306g37f888b7s3a0ede10709f11c3@mail.gmail.com>

On 7/14/06, Steffen Viken Valv?g <steffenv at cs.uit.no> wrote:
> Hi,
>
> First of all, thank you for the time and effort spent developing the
> disksim software.
>
> I recently downloaded disksim 3.0 and attempted to compile it on FreeBSD
> 5.4 with gcc 3.4.2.  I encountered a few minor issues that could easily
> be corrected, and I think these changes could improve portability in
> general.
>
> Attached is a patch file with the changes I made in order to compile;
> there were three instances of labels at the end of compound statements,
> and one issue where a \n was not sufficiently quoted in a perl script
> generating C code.

Thanks for the patch!  In fact, this has been fixed in our CVS tree
for some time.  The upcoming release will include fixes to these
problems but your patch should be useful to other users who've gotten bit by
this issue in the meantime.



john


From chemical.aniket at gmail.com  Tue Jul 18 16:39:02 2006
From: chemical.aniket at gmail.com (Aniket Kulkarni)
Date: Wed, 19 Jul 2006 02:09:02 +0530
Subject: [Disksim-users] what is event simulation
Message-ID: <c6fab9cb0607181339h4e5820fgfde8272d8e5ec654@mail.gmail.com>

Please tell me, what is event simulation?
aniket
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.andrew.cmu.edu/pipermail/disksim-users/attachments/20060719/1f1ec0a1/attachment.html>

From chemical.aniket at gmail.com  Wed Jul 19 11:38:41 2006
From: chemical.aniket at gmail.com (Aniket Kulkarni)
Date: Wed, 19 Jul 2006 21:08:41 +0530
Subject: [Disksim-users] how trace can extract from my system
Message-ID: <c6fab9cb0607190838s86e4c8ch15038162e99f05c5@mail.gmail.com>

Hi,
I have big problem regarding with trace extraction.
I had tried to install ll_trace.
But I can't install.
Tell me any other way to install.
Other thing is that,  which are the parameters in the simulator?
I have figure out some parameters such as
1.block size
2.data layout
3.contiguous memory
How can I vary these parameters?
please tell me as early as possible.
I am waiting the reply.
aniket
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.andrew.cmu.edu/pipermail/disksim-users/attachments/20060719/ff6071e0/attachment.html>

From mbhad001 at cs.fiu.edu  Wed Jul 19 13:19:07 2006
From: mbhad001 at cs.fiu.edu (Medha Bhadkamkar)
Date: Wed, 19 Jul 2006 13:19:07 -0400 (EDT)
Subject: [Disksim-users] About disk models
Message-ID: <33402.131.94.133.71.1153329547.squirrel@www.cs.fiu.edu>


The disk models in Disksim 3.0 are quite old. Are there any validated   
models for larger disks available?

Thanks,
Medha



From mbhad001 at cs.fiu.edu  Wed Jul 19 13:19:19 2006
From: mbhad001 at cs.fiu.edu (Medha Bhadkamkar)
Date: Wed, 19 Jul 2006 13:19:19 -0400 (EDT)
Subject: [Disksim-users] About disk models
Message-ID: <33403.131.94.133.71.1153329559.squirrel@www.cs.fiu.edu>


The disk models in Disksim 3.0 are quite old. Are there any validated   
models for larger disks available?

Thanks,
Medha