Condor and R

classic Classic list List threaded Threaded
3 messages Options
Reply | Threaded
Open this post in threaded view
|

Condor and R

David Reitter
Hi,
I was wondering if anyone has successfully linked R against the  
Condor libraries so that R can be run as a Condor job in the  
"standard" (not "vanilla") universe. The advantage of this would be  
that due to checkpointing, jobs can be suspended and transferred to  
another node. There is a good overview by Xianhong Xie here:

http://cran.r-project.org/doc/Rnews/Rnews_2005-2.pdf

Unfortunately, the article points out that some restrictions of  
Condor preclude us from running R in the standard universe. Does  
anyone know if this can be overcome?

Thanks
David

--
David Reitter - ICCS/HCRC, Informatics, University of Edinburgh

______________________________________________
[hidden email] mailing list
https://stat.ethz.ch/mailman/listinfo/r-help
PLEASE do read the posting guide! http://www.R-project.org/posting-guide.html
Reply | Threaded
Open this post in threaded view
|

Re: Condor and R

Douglas Bates
On 1/24/06, David Reitter <[hidden email]> wrote:

> Hi,
> I was wondering if anyone has successfully linked R against the
> Condor libraries so that R can be run as a Condor job in the
> "standard" (not "vanilla") universe. The advantage of this would be
> that due to checkpointing, jobs can be suspended and transferred to
> another node. There is a good overview by Xianhong Xie here:
>
> http://cran.r-project.org/doc/Rnews/Rnews_2005-2.pdf
>
> Unfortunately, the article points out that some restrictions of
> Condor preclude us from running R in the standard universe. Does
> anyone know if this can be overcome?

I don't think it can currently be overcome and this is not likely to
change soon.  Condor doesn't like dynamic linking and loading for
programs in the standard universe and R more-or-less requires that.
We reached the conclusion that changing either system to be compatible
with the other would require a lot of work.

Mike Redmond <[hidden email]> may have more current information
about the possibility of running R in the standard Condor universe.

______________________________________________
[hidden email] mailing list
https://stat.ethz.ch/mailman/listinfo/r-help
PLEASE do read the posting guide! http://www.R-project.org/posting-guide.html
Reply | Threaded
Open this post in threaded view
|

Re: Condor and R

Michael Redmond
David,

We have discussed proposing a joint project with the UW-Madison Condor
group to develop capabilities in Condor and R that would allow a subset
of batch applications of R to run in the Condor standard universe. To
this point, we have not identified a funding source to support the
development, so the project is on hold.

I am interested in identifying the potential impact of this type of
development as justification in our proposal. If it would affect a large
community of R users, there would be a better possibility to find the
needed funding.

If there is a community of R users out there that would be interested in
having R run in Condor standard universe, please direct comments my way
and, based on interest, I will try to move the proposal forward.

Thanks
Mike Redmond
Director of Computing Services
Dept. of Statistics, UW-Madison
---
Douglas Bates wrote:

> On 1/24/06, David Reitter <[hidden email]> wrote:
>
>>Hi,
>>I was wondering if anyone has successfully linked R against the
>>Condor libraries so that R can be run as a Condor job in the
>>"standard" (not "vanilla") universe. The advantage of this would be
>>that due to checkpointing, jobs can be suspended and transferred to
>>another node. There is a good overview by Xianhong Xie here:
>>
>>http://cran.r-project.org/doc/Rnews/Rnews_2005-2.pdf
>>
>>Unfortunately, the article points out that some restrictions of
>>Condor preclude us from running R in the standard universe. Does
>>anyone know if this can be overcome?
>
>
> I don't think it can currently be overcome and this is not likely to
> change soon.  Condor doesn't like dynamic linking and loading for
> programs in the standard universe and R more-or-less requires that.
> We reached the conclusion that changing either system to be compatible
> with the other would require a lot of work.
>
> Mike Redmond <[hidden email]> may have more current information
> about the possibility of running R in the standard Condor universe.

______________________________________________
[hidden email] mailing list
https://stat.ethz.ch/mailman/listinfo/r-help
PLEASE do read the posting guide! http://www.R-project.org/posting-guide.html