Re: [Condor-users] Condor Benchmarking on Windows Platform

[Steven Timm <timm@xxxxxxxx>]

One thing I can say is that there is a huge amount of
jitter in the condor benchmarking. I have a pool of 410
machines of identical hardware and the MIPS values range
everywhere from 6078 to 8007, with the median at about 6900.
Likewise the kflops range from 1352435 to 1697415.
PArt of it has to do with how busy the machine is otherwise when
it does the benchmarking.

Steve


On Mon, 9 Jun 2008, Dan Bradley wrote:

> Hi James,
>
> The Condor source code for version 7.0 and beyond is fully available on
> the downloads page.  The benchmarking code hasn't changed since 6.8.6.
>
> For calculating MIPS, Condor calls sysapi_mps(), which uses the
> dhrystone benchmark 2.1.
>
> For calculating KFLOPS, Condor calls sysapi_kflops(), which uses
> Linpack.  From a quick glance at the code, I can't tell which version it
> is.  In the Condor source, it's in the file src/condor_sysapi/clinpack.c.
>
> --Dan
>
> James Osborne wrote:
>
> > Hi All
> >
> > One of my users has asked me an interesting question, to which I do
> > not know the answer, so here it is in a nutshell...
> >
> > How are Vanilla Universe KFLOPS and MIPS calculated by Condor version
> > 6.8.6 on the Windows platform ?
> >
> > He wants to perform a comparison between an "average" node in our
> > Condor pool and a regular (non Condor) machine.
> >
> > After gathering the values from various ClassAds he is of the opinion
> > that the values for KFLOPS and MIPS are too high.
> >
> > I presume that for the Java Universe scimark2 from
> > http://math.nist.gov/scimark2 is used (found the jar file)
> >
> > Unfortunately he does not want JavaMIPS, so the question remains...
> >
> > How are Vanilla Universe KFLOPS and MIPS calculated by Condor version
> > 6.8.6 on the Windows platform ?
> >
> > http://www.cs.wisc.edu/condor/manual/v6.8/2_5Submitting_Job.html is a
> > helpful pointer, however...
> >
> > Can you provide as much detail as possible please, so that he can
> >
> > 1) Validate that the KFLOPS and MIPS values advertised in the ClassAd
> > are sensible
> >
> > 2) Run the same benchmark on a non Condor machine
> >
> > KFLOPS - (if Linpack is used , then we will need the version of
> > Linpack used, build instructions or binaries, compiler and version
> > used to build if not binaries, the HPL.dat file used etc)
> >
> > MIPS - (similarly for the Dhrystone benchmark re things needed above)
> >
> > If all of this is integrated into one or more of the Condor daemons
> > (one presumes so), would it be possible to have access to the
> > appropriate source code files ?
> >
> > I suspect Todd or Dan will know who to ask...
> >
> > Thanks in advance
> >
> > Best regards
> >
> > James
> > _________________________________________
> >
> > Dr James Osborne
> > Application Support Engineer
> > Advanced Research Computing @ Cardiff (ARCCA)
> >
> > Cardiff University
> > Redwood Building
> > King Edward VII Avenue
> > Cardiff CF10 3NB
> >
> > Tel +44(0)29 2087 4657
> > Email osborneja1@xxxxxxxxxxxxx
> > www.cardiff.ac.uk/arcca
> > _________________________________________
> >
> > ------------------------------------------------------------------------
> >
> > _______________________________________________
> > Condor-users mailing list
> > To unsubscribe, send a message to condor-users-request@xxxxxxxxxxx with a
> > subject: Unsubscribe
> > You can also unsubscribe by visiting
> > https://lists.cs.wisc.edu/mailman/listinfo/condor-users
> >
> > The archives can be found at:
> > https://lists.cs.wisc.edu/archive/condor-users/
> _______________________________________________
> Condor-users mailing list
> To unsubscribe, send a message to condor-users-request@xxxxxxxxxxx with a
> subject: Unsubscribe
> You can also unsubscribe by visiting
> https://lists.cs.wisc.edu/mailman/listinfo/condor-users
>
> The archives can be found at:
> https://lists.cs.wisc.edu/archive/condor-users/

--
------------------------------------------------------------------
Steven C. Timm, Ph.D  (630) 840-8525
timm@xxxxxxxx  http://home.fnal.gov/~timm/
Fermilab Computing Division, Scientific Computing Facilities,
Grid Facilities Department, FermiGrid Services Group, Assistant Group Leader.