[Bioclusters] job queueing systems for ibm blade center

Rayson Ho bioclusters@bioinformatics.org
Sun, 26 Sep 2004 19:33:49 -0700 (PDT)


While most clusters use either SGE or LSF, it is interesting to note
that Micron uses condor in their compute farms:

http://www.cs.wisc.edu/condor/CondorWeek2003/presentations/gore_micron.ppt



Strong selling point for Condor:
- designed for high-throughput
- powerful checkpointing library (LSF's checkpoint lib was based on it,

  and most SGE clusters use it as well - in the standalone mode)
- free, and the mailing list is very popular (almost as high traffic as

  SGE's)
- follows DRMAA, so the cluster-aware applications you write for SGE 
  will be able to run on Condor with just a recompile.

Not so good:
- less traditional, need to spend time to get used to it
- not opensource, they have been talking about opensourcing it for 
  years :(
- no *BSD ports, and I think x86 Solaris port is missing
- not much commerical support

Rayson


--- Joe Landman <landman@scalableinformatics.com> wrote:
> Condor is a different animal.  It places your code into various 
> environments that you need to link into your code, or into a standard
> 
> environment if the code cannot be relinked. 
> 
> Condor is also less of a "product" and more of a research and 
> development platform.  There aren't too many commercial entities that
> 
> support Condor, while SGE has multiple commercial support options,
> and 
> most hardware OEMs support LSF.
> 
> Condor does run on windows, linux, unix, and other OSes. SGE runs on 
> linux and unix, and windows ports may be in the works.  LSF runs on
> most 
> platforms.  All will enable you to run MPI jobs, run on a grid 
> environment (Condor and SGE have Globus integration as open source 
> options, and Platform has an integrated product suite to enable
> this).
> 
> The run scripts are not interchangeable, and there is a significant 
> learning curve in transitioning from SGE/LSF to Condor (or vice
> versa). 
> 
> Condor is designed more for high throughput computing than classical 
> supercomputing queuing (the basis for LSF and Codine which was the 
> predecessor to GridEngine).  It supports some features such as 
> checkpoint quite nicely (relative to the other environments). It has 
> support for things like cycle stealing and other interesting bits
> that 
> the others do not do.
> 
> As an active research platform, there are other interesting active 
> research tools that surround it. 
> 
> 
> 
> (shameless plug: we support SGE/LSF currently, and may be supporting 
> Condor for a customer shortly)
> 
> 
> Barak Shenhav wrote:
> 
> >Would anyone comment where Condor stands compare to SGE, Torque and
> LSF?
> >
> >On Sun, September 26, 2004 8:06 pm, Chris Dwan said:
> >  
> >
> >>>SGE supports more platforms than TORQUE, but if you only use
> Linux,
> >>>then both of them work fine.
> >>>
> >>>Since SGE/TORQUE are free, you can always try them. If neither of
> them
> >>>work, then you can go for LSF, but it is not free and it supports
> less
> >>>platforms than SGE.
> >>>      
> >>>
> >>LSF: 
> http://www.platform.com/services/support/services/platforms6.0.asp
> >>SGE: 
> http://gridengine.sunsource.net/project/gridengine/download60.html
> >>
> >>How are you counting a "platform?"  It looks like exactly the
> opposite
> >>to me, based on a casual perusal of the lists.
> >>
> >>-Chris Dwan
> >>
> >>_______________________________________________
> >>Bioclusters maillist  -  Bioclusters@bioinformatics.org
> >>https://bioinformatics.org/mailman/listinfo/bioclusters
> >>
> >>    
> >>
> >
> >
> >====================================
> >Barak Shenhav
> >Mayer Bldg, Rm 412
> >Department of Molecular Genetics
> >Weizmann Institute of Science
> >
> >+972  8 9343098 (office)
> >+972  8 9344487 (fax)
> >+972 52 2955550 (cellular)
> >
> >barak.shenhav@weizmann.ac.il
> >http://ool.weizmann.ac.il/
> >
> >  
> >
> 
> -- 
> Joseph Landman, Ph.D
> Scalable Informatics LLC,
> email: landman@scalableinformatics.com
> web  : http://www.scalableinformatics.com
> phone: +1 734 612 4615
> 
> _______________________________________________
> Bioclusters maillist  -  Bioclusters@bioinformatics.org
> https://bioinformatics.org/mailman/listinfo/bioclusters
> 



		
_______________________________
Do you Yahoo!?
Declare Yourself - Register online to vote today!
http://vote.yahoo.com