Sergio, With the need to address the relationship between resources and services I had something like the attached in mind. I don't want to argue about individual attributes yet. It may be useful to associate an attribute with the service to service relationship - this may become evident from the use cases. This will allow you to not only say that a service is related but how it is related. I then have the CE as a service and a cluster as a resource. I suspect that we will want to mimimise the set of resources related to a service - I don't think you want to see every worker node directly. With this view a site is simply a provider of services. Steve
-----Original Message----- From: Sergio Andreozzi [mailto:sergio.andreozzi@cnaf.infn.it] Sent: 03 May 2007 12:10 To: Fisher, SM (Steve) Cc: glue-wg@ogf.org Subject: Re: [glue-wg] GLUE WG phoneconf - 3 May - 4 PM CEST
Hi Steve,
Fisher, SM (Steve) wrote:
Sergio,
I can attend at this time for one hour - if anybody cares to tell me how to phone in.
I will send the details to the list shortly.
I am supposed to be responsible for the service stuff in GLUE and that the schema proposed here is a step back from what we have now and will mess up what we are trying to do with service discovery in SAGA. It does not "incorporate the concepts of GLUE 1.3" I had missed the document when it was first announced on the 3rd April as the subject said "focusing on the CE".
What do you suggest, should I propose an alternative? I don't want to start the public discussion at OGF20 with your definition of service.
As part of the OGF20 meetings we will probably have to revisit the issue of what we mean by optional parameters. This discussion dried up a while ago but it was never resolved. I still insist that the logical modelling has nothing to do with what the implementers/sys-admins may choose to publish.
consider that the document is an early draft. The main attempt was to merge GLUE 1.3 and NorduGrid schema into a common proposal with no backwards-compatible constraints. The main focus was the computing element for which me and Balazs are responsible.
As regards the Core schema, this was put just as a placeholder and is not yet a precise merge of the two (e.g., service data is missing). You have complete freedom of changing it in order to make it aligned with SAGA view before entering the real discussion.
Together with the .png files you can find the UML source code for the ArgoUML software.
Cheers, Sergio
