Andre Merzky a écrit :
Hi all,
Hi Andre,
as most of you should now by now, we plan a SAGA interop demo for OGF30 in Brussels (http://www.ogf.org/OGF30/). In order to pull that off successfully, we need to start to define
- participants - scope - implementations to be used - interop demo scenario
From the top of my head, I would think that the following parties might be interested to team up resources (endpoints, code, people) to get things in place:
- VU Amsterdam - IN2P3, France
Yes, we are interested in participating.
- CCT, LSU - KEK/Naregi - RAL, UK
It would be great if each party could let us know explicitely if they are interested in participating! Did I miss anybody?
We have the following bits and pieces which may, or may not, play a role in the interop demo
implementations: JavaSAGA JSAGA SAGA-C++ PySAGA (over JavaSAGA) PySAGA (over JSAGA) SAGA-Python (over SAGA-C++) command line tools for most of the implementations
functionality: Job Submission (all) Data transfer / access (all) Advert Service (not JSAGA I think, not in PySAGA)
Not JSAGA indeed.
Replica Management (not JSAGA?)
JSAGA supports Replica Management package for gLite-LFC, iRODS and SRB.
Service Discovert (not in JSAGA? not in PySAGA)
Not yet in JSAGA.
backends: local (all) globus (all) ssh (all)
OK.
aws (all?)
Not in JSAGA.
glite (all?)
OK for us.
bes (all?)
Not in JSAGA.
Infrastructures local institutions teragrid loni naregi What about European Grids?
Again, is there something I miss?
PySAGA appears to be a great integration point, and SAGA-C++ intents to support it very soon, too - but it is not sure that we manage to do that before OGF30.
A simple interop demo would be to submit the same job (NOT /bin/date) to a set of resources in the various infrastructures discovered via SD, from various tools. A job submitted via python for example should be monitorable from C++ tools, and output could be reaped via PySAGA-over-JSAGA, etc.
This example will not work with some adaptors because JSAGA adaptors sometimes add in job description some information that is not be used by the middleware, but used by the adaptor in order to overcome some limitations in middleware features, in particular for reaping output. However, if we restrict to the features that are natively supported by the middleware (e.g. save output on gsiftp for gLite), it should work.
The above is just an initial input, to get the discussion and planning started. Please feed back, and complete the item lists above. Once we have those lists complete, we should be able to come up with some more or less realistic scenario.
I'll mirror thiss list on our wiki at GridForge, so that we can edit things in place. Feel free to discuss on the list though, I'll try to keep the thread in sytnc with the wiki.
Best, Andre.
Best regards, Sylvain