суббота, 16 мая 2009 г.

Sage on BalticGrid

A presentation was made on the BalticGrid 2nd All Hands Meeting in Riga, Latvia on 12-14 May 2009.
As the result Sage was recommended for BalticGrid gLite-sites as a service that could be installed by all administrators. You can follow the progress of SAGE installations on http://infosite.balticgrid.org/ looking up for SAGE in software environment.
The chosen tools from RESPECT program were confirmed to be appropriate for the task of SAGE's gridification.
Many of the colleagues have already used Sage for their purposes and the proposal to use Sage on Grid was accepted at once.

The next thing to be done is to create an rpm package for automatic installation of SAGE on SL4.7 disabled of web server functionality.
More about the first TODOs is posted to the sage-devel mailing list here:

7 комментариев:

  1. Great slides! The High Energy Physics community would benefit a lot from python/sage in grid computing. As one can see at the moment there is lots of interfacing of old C, Fortran libraries with python analysis tools going on, due to faster development cycles.
    Are there any examples of using the RESPECT tools with standard python? Maybe there is not even that much customization necessary.

  2. The truth is that I really don't know what kind of customization is necessary. I hope to catch up things in the process of integration of Sage with glite-UI, DIANE and Ganga.
    At first, obviously, a separate sage function is needed to deal with underlying tools. Then all glite, ganga, diane executables should also be accessable from ipython.
    I didn't find any examples to use these tools with python yet because I just started out.
    About interface to C, Fortran: the great thing is that it is possible to use old code in C and Fortran under python interfaces such as Cython and ... (I would appreciate if someone could point me to the same tool for Fortran).
    I hope I'll manage to produce smth valueable.

  3. I've just found out that there is a plan for a Fortran Support in Cython.
    The goal is to allow transparently call Fortran code from Cython.
    See here http://wiki.cython.org/enhancements/fortran

  4. This support makes Sage a profound computing framework for existing C/C++ and Fortran libraries.

    There is also "The F2PY Project" http://www.f2py.org/

    Sage is definitely the best bet.

  5. Cython support for Fortran on the same scale as for C would be a breakthrough.. f2py is already functional but not the tool for high speed computing. Anyhow If I was you I would not start to try packing a vast beast like sage. Maybe providing the tools "glite-UI, DIANE and Ganga" in a sage grid.spkg package would be a first solution. (This way I could test them on our machines)

    However, once the tools are accessible at the same time as sage, one could try to run some test computations (e.g. just sum up some numbers) and document the basic approach.

    Also there are some grid relevant things already available for sage ("mpi4py"..)

  6. Yeah, I'm already working in the direction of creating a standalone grid package for Sage.
    This way seems to be the right one.
    I also need to do some testing, you are right.

    I can package Sage with DIANE and Ganga but glite-UI seems to be very complex to pack into Sage because it needs additional complex "tuning" and tar version of UI basically is outdated.

    Now I'm working on the first part of creating Sage-WN distribution (see http://groups.google.com/group/sage-devel/browse_thread/thread/b42526c61b74daa5 )

    I haven't worked with mpi4py. Sure it also can be somehow useful in the Grid environment.

  7. Have you considered building some more features into SPD (http://code.google.com/p/qsnake/wiki/SPD) insteatd of ripping things out of sage? At least to create the WN software this could be the more secure solution.