(Reposting this as my original reply appeared in its own thread)
As Drew pointed out, filebench is included in /usr/benchmarks already
in OpenSolaris (and if its not, then you should be able to get it via
IPS).
I just downloaded the source, built it (following the directions found
in the INSTALL
> Further, the package build from SourceForge doesn't install properly on
> OpenSolaris 2008.11 amd64, doesn't run properly on Linux, and doesn't build
> on Mac OS X. I would be happy to provide input and work-arounds, but I can't
> tell if it's worth the trouble, as all activity on filebench ap
Fellow FileBenchers,
Version 1.3.4 is now available at:
http://sourceforge.net/project/showfiles.php?group_id=133644
1.3.4 includes fixes for:
6695801 fileset_pick() lock contention with many threads
6711285 perf regression with (lots of) robust locks since robust locks
got revisited (PSARC 200
On Jun 3, 2008, at 9:26 PM, [EMAIL PROTECTED] wrote:
> On Tue, Jun 03, 2008 at 03:32:47PM -0700, eric kustarz wrote:
>>
>> On May 29, 2008, at 11:44 PM, [EMAIL PROTECTED] wrote:
>>> [...]
>>>
>>> I remember another hardware arrays (yes, two arrays) test
On May 29, 2008, at 11:44 PM, [EMAIL PROTECTED] wrote:
> On Thu, May 29, 2008 at 02:04:21PM -0700, eric kustarz wrote:
>>>
>>> Currently both tools gives different view of io. While filebench
>>> simulates real workload, it cannot show what vdbench shows.
>&g
On May 30, 2008, at 7:56 AM, Sean McGrath - Sun Microsystems Ireland
wrote:
> eric kustarz stated:
> < >
> < > Currently both tools gives different view of io. While filebench
> < > simulates real workload, it cannot show what vdbench shows.
> < > E.g.
>
> Currently both tools gives different view of io. While filebench
> simulates real workload, it cannot show what vdbench shows.
> E.g. we were doing some hardware array tests and it turned out (using
> vdbench) that the array works in a really strange way: sometimes its
> IOPS jumps very high, s
fellow filebenchers,
Version 1.3.1 is now available at:
http://sourceforge.net/project/showfiles.php?group_id=133644
Thanks to Denis' changes, FileBench is once again buildable on Linux:
http://www.opensolaris.org/jive/thread.jspa?threadID=57737&tstart=15
In addition, I made some minor changes t
fellow filebenchers,
Version 1.3.0 is now available at:
http://sourceforge.net/project/showfiles.php?group_id=133644
Version 1.3.0 adds support for composite flowops as detailed here:
http://www.opensolaris.org/jive/thread.jspa?threadID=58827&tstart=0
http://www.solarisinternals.com/wiki/images/3
>
>>> 3. Please give some more description about every part of the source
>>> tree, which provided on the sourceforge?
>> There is a fairly detailed description of how filebench works on
>> SolarisInternals.com. Also documentation on the "f" language is
>> there.
> you mean the wiki documentation
Instead of 'lex', try using 'flex'. I was able to build successfully with:
fsh-suzuki% flex --version
flex version 2.5.4
fsh-suzuki%
fsh-suzuki% make package
...
flex -t parser_lex.l > parser_lex.c
source='parser_lex.c' object='parser_lex.o' libtool=no \
depfile='.deps/parser_lex.Po' tmpdepfile=
Fellow filebenchers,
Version 1.2.1 is now available at:
http://sourceforge.net/project/showfiles.php?group_id=133644
This add's Drew's random variables as mentioned at:
http://www.opensolaris.org/jive/thread.jspa?messageID=190041𮙙
http://www.solarisinternals.com/wiki/images/1/1d/
FileBench_Rando
On Feb 28, 2008, at 1:19 PM, Jonathan Edwards wrote:
>
> On Feb 28, 2008, at 1:39 PM, eric kustarz wrote:
>> Fellow filebenchers,
>>
>> Version 1.1.1 is now available at:
>> http://sourceforge.net/project/showfiles.php?group_id=133644
>
> thanks eric ..
>
Fellow filebenchers,
Version 1.1.1 is now available at:
http://sourceforge.net/project/showfiles.php?group_id=133644
This adds Drew's "run to completion" mode as detailed at:
http://www.opensolaris.org/jive/thread.jspa?threadID=52153&tstart=0
With this, life is much nicer for workloads that are
Excellent news!
Yes, indeed, let's work together to get a stable and complete code drop that
fully works on both OpenSolaris and Linux.
eric
This message posted from opensolaris.org
___
perf-discuss mailing list
perf-discuss@opensolaris.org
I've updated the build infrastructure so it all should work (on Solaris10 as
well as Nevada):
http://blogs.sun.com/erickustarz/entry/filebench_1_1_0_complete
There are even pre-built packages of the latest version of FileBench (1.1.0)
(that work on s10 and Nevada).
Let me know if you run into a
I'm happy to report that we have updated the building process on the
sourceforge site so that anyone can do a complete build with the
updated source (as in the freshest FileBench source in the world).
This complete build not only builds the open source parts
(go_filebench, filebench.pl, wor
Yep, these changes look great.
Besides the very nice merge of files into filesets and others, a special thanks
for fixing up oltp.f.
eric
This message posted from opensolaris.org
___
perf-discuss mailing list
perf-discuss@opensolaris.org
I've filed:
6624946 filemicro_seqread.f uses fileset instead of file
to track this.
This message posted from opensolaris.org
___
perf-discuss mailing list
perf-discuss@opensolaris.org
hey Marion,
If you need just the binaries for 1.0.0, then you can either simply upgrade to
the latest ON build (76+) or if you download the source and build it, then
you'll create the package (SUNWfilebench) yourself. Sorry for the delay, but
we do have someone cleaning up the Makefiles/puttin
Isn't this a bug in the workload definition?
filemicro_seqread.f specifies the file you're sequentially reading as a fileset:
"
define fileset
name=bigfileset,path=$dir,size=$filesize,entries=$nthreads,dirwidth=1024,prealloc=100,cached=$cached
"
I believe that should be a 'file' - much like sing
I would guess that you aren't giving FileBench enough work to do, see:
http://www.opensolaris.org/os/community/performance/filebench/gotchas/
http://www.solarisinternals.com/wiki/index.php/FileBench#Recommended_parameters
eric
This message posted from opensolaris.org
__
On Oct 6, 2007, at 12:10 PM, Alan Coopersmith wrote:
> eric kustarz wrote:
>> When the SFW gate can handle CTF then we can talk. Until then ON
>> is the only way to go for what we wanted.
>
> No gate outside ON can handle CTF until ON decides to make it
> a
When the SFW gate can handle CTF then we can talk. Until then ON is the only
way to go for what we wanted.
eric
This message posted from opensolaris.org
___
perf-discuss mailing list
perf-discuss@opensolaris.org
On Aug 29, 2007, at 9:16 PM, Peter C. Norton wrote:
> On Wed, Aug 29, 2007 at 09:55:24AM -0700, eric kustarz wrote:
>> No official way to do it now, but something i would definitely
>> like to see happen.
>>
>> Another hackish way is to run multiple instances at th
No official way to do it now, but something i would definitely like to see
happen.
Another hackish way is to run multiple instances at the same time (each with
its own $dir specified). Coordinating the separate instances and aggregating
the data would be tricky.
Part of this ties into getting
Cool.
Yeah we'll definitely take a look at your suggested fix. Just need to see if
it makes more sense/fits the framework to pass in a mode argument instead of
dup'ing functionality. We'll look to see if we could do something akin to:
flowop openfile name=openfile1,filesetname=bigfileset,f
Once the FileBench ARC case finishes up, then we'll update sourceforge with our
changes. After that, we'll finish up the process work of getting FileBench
into OpenSolaris.
The ARC case should be done within a week. You can track it here:
http://www.opensolaris.org/os/community/arc/caselog/200
>
> Message: 1
> Date: Thu, 26 Jul 2007 12:28:43 -0400
> From: [EMAIL PROTECTED]
> Subject: [perf-discuss] filebench on linux
> To: perf-discuss@opensolaris.org
> Message-ID: <[EMAIL PROTECTED]>
> Content-Type: text/plain; charset=ISO-8859-1; format="flowed"
>
> Hello,
>
> I was able to get
Did you install filebench as a package?
What does 'pkginfo -l filebench' return?
Have you changed the default location of where 'filebench' resides? The
version you're using should be expecting 'filebench' to live in
/opt/filebench/bin/filebench.
eric
This message posted from opensolaris.
hiyah bruce,
Yeah looks like filebench is being lazy and not using workload specifics on
whether or not to open the file read only or rw.
Out of curiosity, is the delegation problem (what should be unnecesary) recalls?
I'll track that bug...
(and yep, there appears to be a delay on when email
Hey Tim,
Funny you should ask. We are actually heavily in the midst of updating
filebench. Part of that will be integrating it into ON (yes we will kick back
all changes to sourceforge). Spencer is heading up the work to create a
NFSv3/v4 plug-in very much along the lines of what you are thi
Is there a way to set the seed value for random IO workloads in filebench (so
that i can fairly compare random IO runs)?
eric
This message posted from opensolaris.org
___
perf-discuss mailing list
perf-discuss@opensolaris.org
thanks richard
and in case the suspense was killing you, here's what i originally posted to a
Sun internal list...
"
hey fellow benchers,
I was looking for a workload to do random reads and writes at the same time...
didn't see one, so i merged the multistreamread.f and randomread.f workloads:
hey fellow benchers,
I was looking for a workload to do random reads and writes at the same
time... didn't see one, so i merged the multistreamread.f and
randomread.f workloads:
#
# Copyright 2005 Sun Microsystems, Inc. All rights reserved.
#
# The contents of this file are subject to the te
35 matches
Mail list logo