Hi
I'm not sending this to Bob, Erwin and Heinz, because although I'm
really angry I suspect that swearing profusely at top level EGEE
management is probably unwise - they do however deserve it...
On 31 Oct 2007, at 23:16, David Colling wrote:
> Dear All,
>
> I mailed Heinz with the concerns expressed here and on the dteam
> list about him running RSA-768 challenge jobs under the biomed VO
> and this being seen as abuse. I also asked about the size of the
> jobs being a problem.
>
> Dealing with the second point first...
> Heinz' reply was that he did put the size of the job in the JDL so
> it was only running at sites that were advertising that they could
> cope with such jobs - as ours must have been (Mona and Kostas,
> please can you look at this for our site as clearly we cannot cope
> with more than 2 per 4 core machine).
>
> On the first point...
> (Heinz, please correct me if I am wrong.)
>
> Heinz said that he was indeed running as part of this challenge and
> was doing so with the full encouragement of Bob and Erwinwho were
> very keen to see the results. He was doing this as biomed because
> of a lack of any other VO that was any more appropriate.
From the CIC portal, biomed described itself as:
"These VO covers the areas related to health sciences. Currently, it
is divided in 3 sectors: medical imaging, bioinformatics and drug
discovery."
We support the VO for it to engage in _that_ work, and we're happy to
have done work related to malaria, avian flu, etc. However, I don't
see anything about rsa768 factorisation.
So, this is, to my mind, even worse. This is not just Heinz being a
loose cannon, but sites being conned by top level EGEE management
into running jobs to which they had in no way agreed to run.
The problem was then exacerbated by the way that Heinz wrote the
code, which resulted in biomed being able to grab far more of many,
many clusters in the UK than was reasonable. (And so much for EGEE
promoting push model RBs - just send in the pilots and watch our
fairsharing go all to hell.)
Frankly, as the UK, I think we should give them a bloody rocket for
this. They've shown huge disrespect to sites - and how on earth can
they expect other EGEE users and VOs to play by the rules when then
engage in such a gross violation of our trust?
>
> Explained in these terms I can see all sides of this discussion as
> having some validity. I can see why "the management" are keen
> perform very well in such challenges. Heinz is only a member of the
> biomed VO and besides, currently, there isn't an appropriate VO for
> this work. Equally, I can see that a site like Glasgow (chosen only
> because Graeme has the hard numbers) viewing it as abuse that
> 50,000 CPU hours that they thought were going on biomedical
> research being used on this challenge, especially when they have
> very many queuing Atlas jobs. While they might be very happy to be
> part of this challenge (or may not be I haven't asked) they would
> probably haven given it a lower priority than biomedical research.
> Clearly, the answer is to set up up a challenge VO that people can
> support in its own right for such things (and can give it whatever
> priority they want). Is this something that we should do in the UK
> or should be somewhere else (or doesn't it matter where ;-)
>
> There was the further problem that raising tickets against biomed
> didn't get any response for sometime and then I don't get the
> impression that they had actually made it as far as Heinz. Showing
> a communication issue.
>
> This incident has caused quite a problem, resulting in Heinz/biomed
> being banned from large (and increasing) fraction of UK sites for
> abuse. I would suggest both that this ban be lifted and that a
> challenge VO is set up for this sort of thing sooner rather than
> later.
We haven't banned biomed - we've banned Heinz. And I am in no hurry
to unban him. I'd expect an apology at the very least, as well as an
assurance that this will not happen again.
graeme (who gets rather tetchy after the witching hour...)
--
Dr Graeme Stewart - http://wiki.gridpp.ac.uk/wiki/User:Graeme_stewart
ScotGrid - http://www.scotgrid.ac.uk/ http://scotgrid.blogspot.com/
|