[GRIDCPR-WG] CPR Document - final call
Eduardo Huedo Cuesta
ehuedo at fdi.ucm.es
Thu Mar 29 08:23:58 CDT 2007
Dear All,
From the GridWay team, we would like to propose another consumer
use-case that we think is within the scope of GridRPC. See below.
Best Regards,
Eduardo Huedo.
--------------------------------------------------------------------
GridWay Metascheduler
=================
The GridWay Metascheduler [1, 2], now a Globus project, adapts job
execution to changing grid conditions by providing fault recovery
mechanisms, dynamic scheduling, migration on-request and opportunistic
migration [3]. Migration is implemented by restarting the job on the new
candidate host, therefore the job should generate restart files at
regular intervals in order to continue execution from a given point. If
checkpointing files are not provided, the job is restarted from the
beginning. GridWay periodically retrieves to the client machine or a
checkpoint server (GridFTP URL) the restart architecture-independent files.
Jobs submitted with GridWay could benefit from GridCPR systems providing
standard and uniform APIs and services for portable checkpoint
generation and storage.
Functional requirements
. API for application state writing and reading.
. Services for failure notification.
. Services for checkpoint data management.
[1] GridWay Metascheduler. http://www.gridway.org/.
[2] E. Huedo, R.S. Montero and I.M. Llorente: A framework for adaptive
execution on grids. Software - Practice and Experience 34 (7): 631-651,
2004.
[3] E. Huedo, R. S. Montero, I. M. Llorente: Evaluating the reliability
of computational grids from the end user's point of view. Journal of
Systems Architecture 52(12): 727-736, 2006.
Andre Merzky escribió:
> Hi groups,
>
> as discussed earlier, we put some effort into the GridCPR
> documents, to get them back into the editor pipeline.
> Thanks to Nathan and others, both the CPR usecase and the
> cpr architecture document have now all public comments
> addressed, and are to be submitted to the OGF editor.
>
> The docs are supposed to represent groups consensus after
> submission, so, this mail is a one week final call on the
> mailing list: please review the documents, and comment on
> them! "speak now or forever hold your peace ..." :-)
>
> Cheers, Andre.
>
>
> ------------------------------------------------------------------------
>
> --
> gridcpr-wg mailing list
> gridcpr-wg at ogf.org
> http://www.ogf.org/mailman/listinfo/gridcpr-wg
--
GridWay, Meta-scheduling Technologies for the Grid! http://www.gridway.org
**************************************************
Dr. Eduardo Huedo Cuesta
Departamento de Arquitectura de Computadores
y Automática
Facultad de Informática
Universidad Complutense de Madrid
C/ Prof. García Santesmases s/n
28040 Madrid
Spain
Tel: +34 91 394 76 03
Fax: +34 91 394 75 27
Email: ehuedo at fdi.ucm.es
**************************************************
More information about the gridcpr-wg
mailing list