[Swift-devel] Re: poor performance

[Michael Wilde]

Its likely that I lost track of this.

I *thought* that when Marc reported the bug, that he had already applied the fix, but now that you remind me of the thread, its very likely that he did not.

I will do this with him (move him to 0.92 RC), change the throttle values, and re-test the workflow.

- Mike


----- Original Message -----
> I thought I mentioned what the problem was there and that it should be
> fixed in cog 4.1.8/swift 0.92.
> 
> Mihael
> 
> On Tue, 2011-01-11 at 12:07 -0600, Michael Wilde wrote:
> > Mihael, I moved the log below to ~wilde/ftdock on the CI net.
> >
> > I tried to run swift-plot-log on it, and that failed ( i got
> > nothing).
> > I'll file the latter issue in bugzilla.
> >
> > - Mike
> >
> > ----- Original Message -----
> > > Mihael, can you take a look at Marc's logs? Do you have access to
> > > engage-submit?
> > >
> > > We should now help Marc move his work to bridled and communicado
> > > where
> > > we now have COndor-G installed and can more readily assist him in
> > > debugging.
> > >
> > > Marc, I'll try to give this more attention this week, but have to
> > > work
> > > on a deadline for today, first.
> > >
> > > - Mike
> > >
> > >
> > > ----- Forwarded Message -----
> > > From: "Marc Parisien" <mparisien at uchicago.edu>
> > > To: "Michael Wilde" <wilde at mcs.anl.gov>
> > > Sent: Monday, January 10, 2011 10:25:30 AM
> > > Subject: poor performance
> > >
> > > Hi Mike,
> > >
> > > I have a campaign running on renci for almost 2 days now; only 529
> > > on
> > > 3000 jobs are done. IBI can weed 3000 within 2 days on 128
> > > processors.
> > > What is the problem? Why don't I have a decent performance on
> > > making
> > > use of 10 super-computers? I have the most simplest swift script
> > > ever
> > > (aside from the "Hello World" one).
> > >
> > > the (1 Gb) log file is here:
> > > /home/parisien/Database/MCSG/ftdock-20110108-1301-xx8evgh7.log
> > >
> > > could it be because of a bad "site" that throws off all of Swift's
> > > scheduling? Or a "badly" set parameter??
> > >
> > >
> > > Very Best,
> > > Marc.
> > >
> > >
> > > --
> > > Michael Wilde
> > > Computation Institute, University of Chicago
> > > Mathematics and Computer Science Division
> > > Argonne National Laboratory
> >

-- 
Michael Wilde
Computation Institute, University of Chicago
Mathematics and Computer Science Division
Argonne National Laboratory