From dtomas at cose.sfsu.edu  Thu Jun  1 14:27:53 2006
From: dtomas at cose.sfsu.edu (Daniel Tomasevich)
Date: Thu, 01 Jun 2006 11:27:53 -0700
Subject: [Disksim-users] validated disk parameters
Message-ID: <447F31A9.3060303@cose.sfsu.edu>


Hi,

The validated database of disk parameters on G Granger's website under
DiskSim/diskspecs.html was very helpful for research.

Has anyone posted similar parameter extraction data for other disks?

Thanks,
Daniel



From bucy at ece.cmu.edu  Fri Jun  2 14:03:17 2006
From: bucy at ece.cmu.edu (John Bucy)
Date: Fri, 2 Jun 2006 11:03:17 -0700
Subject: Fwd: [Disksim-users] validated disk parameters
In-Reply-To: <4fc1e0430606011153h77d3b07ax144d6be80b0e6219@mail.gmail.com>
References: <447F31A9.3060303@cose.sfsu.edu>
	<4fc1e0430606011153h77d3b07ax144d6be80b0e6219@mail.gmail.com>
Message-ID: <4fc1e0430606021103h7693170fmdccd7c0aa515f563@mail.gmail.com>

On 6/1/06, Daniel Tomasevich <dtomas at cose.sfsu.edu> wrote:
>
> Hi,
>
> The validated database of disk parameters on G Granger's website under
> DiskSim/diskspecs.html was very helpful for research.
>
> Has anyone posted similar parameter extraction data for other disks?

The upcoming release of disksim will include validated models of much
newer disks, e.g. Seagate Cheetah15k4, etc.

If we put up a wiki or something, we could perhaps have a "user
contributed" section for stuff like this.


john


From shahrukh at cs.virginia.edu  Mon Jun 12 10:32:39 2006
From: shahrukh at cs.virginia.edu (Shahrukh Rohinton Tarapore)
Date: Mon, 12 Jun 2006 10:32:39 -0400
Subject: [Disksim-users] disk block numbers
Message-ID: <448D7B07.7080609@cs.virginia.edu>

Is it safe to assume that an array of identical disks will each have the 
exact same number of disk blocks, as reported by getdisk(i)->numblocks?  
I want to make sure there isn't some arbitrary way for one disk to have 
even one or two more blocks than another (although this happens in real 
disks).  Thanks.

-- shahrukh


From shahrukh at cs.virginia.edu  Tue Jun 13 16:42:14 2006
From: shahrukh at cs.virginia.edu (Shahrukh Rohinton Tarapore)
Date: Tue, 13 Jun 2006 16:42:14 -0400
Subject: [Disksim-users] accessing logorg struct
Message-ID: <448F2326.4030805@cs.virginia.edu>

Sorry to ask such an implementation type question but I've been plagued 
by this for almost 2 days.  I am writing a program which controls the 
disksim simulator (i.e. treats it as a slave).  Thus I need certain 
information regarding the state of the simulation.  Specifically I need 
to if the storage system is addressed by Array or by Parts.  It seems 
there is no way to access the logical organization structure (struct 
logorg which contains a member called addrbyparts) from outside the 
disksim code base.  I've delved deeply into the disksim code and I'm not 
even sure how its done in the base code.

I've narrowed my problem to the fact that the disksim_t struct uses 
forward declarations to make the compiler happy when defining pointers 
to other structs like iodriver_info.  The errors I get imply that the 
unit I am compiling cannot see all of the structure definitions it 
needs.  I think the forward declarations allow this to occur.

Hopefully this makes some sense; in my own code I try to access the 
addrbyparts variable like this:
    disksim->iodriver_info->sysorgs->addrbyparts
where disksim, iodriver_info, and sysorgs are structure pointers and 
which are members in the subsequent structure.  Disksim has macros 
defined similiar to this in disksim_iodriver.c, so its odd that it 
wouldn't work.

 I tried a little test:
    int x = disksim->intqlen;  // this compiles fine which means the 
disksim struct definition can be seen

Then I tried:
    int x = disksim->iodriver_info->numiodrivers  // this gives me the 
same error I initially incountered regarding the compliation unit not 
seeing the definition of struct iodriver_info.

Any help resolving this issue would be appreciated.  I hope changing the 
core disksim code is not necessary as I would like my own code to be 
able to work as an extension to disksim rather than as patch work to 
arbitrary code segments.  Thanks.


-- shahrukh


From bucy at ece.cmu.edu  Tue Jun 13 16:59:14 2006
From: bucy at ece.cmu.edu (John Bucy)
Date: Tue, 13 Jun 2006 13:59:14 -0700
Subject: [Disksim-users] accessing logorg struct
In-Reply-To: <4fc1e0430606131354r2b499b9aj90856b45ea8d3911@mail.gmail.com>
References: <448F2326.4030805@cs.virginia.edu>
	<4fc1e0430606131354r2b499b9aj90856b45ea8d3911@mail.gmail.com>
Message-ID: <4fc1e0430606131359q7bc422b8q621cc775f5ed6a96@mail.gmail.com>

On 6/13/06, Shahrukh Rohinton Tarapore <shahrukh at cs.virginia.edu> wrote:
> Sorry to ask such an implementation type question but I've been plagued
> by this for almost 2 days.  I am writing a program which controls the
> disksim simulator (i.e. treats it as a slave).  Thus I need certain
> information regarding the state of the simulation.  Specifically I need
> to if the storage system is addressed by Array or by Parts.  It seems
> there is no way to access the logical organization structure (struct
> logorg which contains a member called addrbyparts) from outside the
> disksim code base.  I've delved deeply into the disksim code and I'm not
> even sure how its done in the base code.

I have never worked extensively in that code so my short answer is, "I
don't know."

We made a point of separating out the interface stuff such that it
doesn't include any other internal disksim headers which are really
messy and will often caused name/symbol clashes with other code.  You
probably need to #include some of these headers in order to reach that
far down into the data structures.

Another option might be to get it "out of band" somehow, e.g. read it
out of the parameter file yourself.



john


From sgovinda at cse.psu.edu  Sun Jun 18 06:38:20 2006
From: sgovinda at cse.psu.edu (Sriram Govindan)
Date: Sun, 18 Jun 2006 06:38:20 -0400 (EDT)
Subject: [Disksim-users] (no subject)
Message-ID: <Pine.SOC.4.64.0606180631400.11004@zoller.cse.psu.edu>

Hi all,

I am a new user of disksim. I have some doubts listed below.

a) Instead of submitting all the disk requests in a trace file, i want to 
send the subsequent requests only when i know that the previous request is 
done with the disk, that is number of outstanding request should be one. 
In short, i want to know if the previous request has finished using the 
disk.

b) How to increase the size of the disk? (say barracuda disk from 2 gb to 
40gb), i understand that it is not straight forward (just increasing the 
number of sectors)

thank you,

sriram.G
Phd student,
Computer science department,
Penn state university.


From sgovinda at cse.psu.edu  Sun Jun 18 07:26:25 2006
From: sgovinda at cse.psu.edu (Sriram Govindan)
Date: Sun, 18 Jun 2006 07:26:25 -0400 (EDT)
Subject: [Disksim-users] run multiple trace files
Message-ID: <Pine.SOC.4.64.0606180723480.11047@zoller.cse.psu.edu>

Hi all,

I want to run multiple trace files (like multiple workloads accessing the 
same disk) simultaneously, each having its own trace files (these are not 
synthetic workloads). I do not know how to go about with this.

thank you,

sriram.G


From bucy at ece.cmu.edu  Wed Jun 21 14:31:16 2006
From: bucy at ece.cmu.edu (John Bucy)
Date: Wed, 21 Jun 2006 11:31:16 -0700
Subject: [Disksim-users] (no subject)
Message-ID: <4fc1e0430606211131v6ef6616dmf13ba43cc23b9338@mail.gmail.com>

On 6/18/06, Sriram Govindan <sgovinda at cse.psu.edu> wrote:
> Hi all,
>
> I am a new user of disksim. I have some doubts listed below.
>
> a) Instead of submitting all the disk requests in a trace file, i want to
> send the subsequent requests only when i know that the previous request is
> done with the disk, that is number of outstanding request should be one.
> In short, i want to know if the previous request has finished using the
> disk.

There is not a way to do this out of the box.  What I would do is use
the external interface to run disksim as a library and use some
lightweight threads package (gnu pth, state threads, etc) and run the
disksim event loop in one and each of your trace replays in separate
threads.  Your runtime would synchronize all of this so that the
simulated IOs in each of the trace replays would look like a blocking
call in simulated time.  Does that make sense?

> b) How to increase the size of the disk? (say barracuda disk from 2 gb to
> 40gb), i understand that it is not straight forward (just increasing the
> number of sectors)

To get anything like realistic results, it is non-trivial to do this
because all of those parameters interact in pretty subtle ways.  Also,
that barracuda model is from ~1994 IIRC -- really really old.  I know
this probably doesn't help you right now but we are going to have a
new release that will include valdiated models of much newer, larger
disks.




john


From shahrukh at cs.virginia.edu  Mon Jun 26 17:47:24 2006
From: shahrukh at cs.virginia.edu (Shahrukh Rohinton Tarapore)
Date: Mon, 26 Jun 2006 17:47:24 -0400
Subject: [Disksim-users] simulation rollback
Message-ID: <44A055EC.6000104@cs.virginia.edu>

How difficult is it rollback a simulation run?  What I mean is if during 
a simulation run the simulated time at 25 ms. and I want it go go back 
to the state the simulator was in at time 20 ms what changes are 
necessary.  Would I only need to remember the state of the intq at time 
20 ms?  Also what about the state of the mechanical model?  Please let 
me know if there are other conciderations to be made and if  rollback is 
even feasible? Thanks

-- shahrukh


From bucy at ece.cmu.edu  Mon Jun 26 17:53:59 2006
From: bucy at ece.cmu.edu (John Bucy)
Date: Mon, 26 Jun 2006 14:53:59 -0700
Subject: [Disksim-users] simulation rollback
In-Reply-To: <4fc1e0430606261453k49563e34j93b5e1b809cb97b@mail.gmail.com>
References: <44A055EC.6000104@cs.virginia.edu>
	<4fc1e0430606261453k49563e34j93b5e1b809cb97b@mail.gmail.com>
Message-ID: <4fc1e0430606261453v60bdec08r673541b59ac7afdc@mail.gmail.com>

On 6/26/06, Shahrukh Rohinton Tarapore <shahrukh at cs.virginia.edu> wrote:
> How difficult is it rollback a simulation run?  What I mean is if during
> a simulation run the simulated time at 25 ms. and I want it go go back
> to the state the simulator was in at time 20 ms what changes are
> necessary.  Would I only need to remember the state of the intq at time
> 20 ms?  Also what about the state of the mechanical model?  Please let
> me know if there are other conciderations to be made and if  rollback is
> even feasible? Thanks

There is some old checkpointing code in there but I don't think its
been used for years and I don't know whether it even works anymore.  I
never used it my self and I'm not sure if its fine-grained enough for
what you're describing here.

Otherwise, I'm not aware of anyone in our group actively looking into
doing this for disksim.  Off the top of my head, I imagine it would be
rather difficult and invasive.



john


From mitzytest at gmail.com  Tue Jun 27 13:08:31 2006
From: mitzytest at gmail.com (Nuengwong Tuaycharoen)
Date: Tue, 27 Jun 2006 13:08:31 -0400
Subject: [Disksim-users] RAID5 and RAID10
Message-ID: <ce68cf5a0606271008j522adc96h48a69c54511af20a@mail.gmail.com>

Hi,
I'm using Disksim 2.0. Where can I find a sample of RAID5 and RAID10
parameter files and diskspec files?

Thanks.
--NT
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.andrew.cmu.edu/pipermail/disksim-users/attachments/20060627/ac7e14e8/attachment.html>

From xyao at cse.unl.edu  Tue Jun 27 13:14:35 2006
From: xyao at cse.unl.edu (Xiaoyu Yao)
Date: Tue, 27 Jun 2006 12:14:35 -0500
Subject: [Disksim-users] RAID5 and RAID10
In-Reply-To: <ce68cf5a0606271008j522adc96h48a69c54511af20a@mail.gmail.com>
Message-ID: <000c01c69a0d$2ccde060$0ab05d81@Safari>

Hi,
 
If you use disksim 3.0, the valid directory has many good samples for RAID
0, 1, 5.  RAID 10 can be configured based on these samples with slight
modification. 
 
xiaoyu

  _____  

From: disksim-users-bounces at ece.cmu.edu
[mailto:disksim-users-bounces at ece.cmu.edu] On Behalf Of Nuengwong
Tuaycharoen
Sent: Tuesday, June 27, 2006 12:09 PM
To: disksim-users at ece.cmu.edu
Subject: [Disksim-users] RAID5 and RAID10


Hi,
I'm using Disksim 2.0. Where can I find a sample of RAID5 and RAID10
parameter files and diskspec files?
 
Thanks.
--NT
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.andrew.cmu.edu/pipermail/disksim-users/attachments/20060627/973c26db/attachment.html>