The 3rd CGS-WG telecon was held this morning (Pacific time). Attendees were:
Larry Flon (USC/ISI)
Jenny Schopf (ANL)
Viktor Mihajlovski (IBM Germany)
Andreas Maier (IBM Germany)
Tom Roney (NCSA)
Eric Woo (IBM Yorktown)
Peter Gietz (DAASI)
We reviewed the meetings at GGF6 in Chicago
last week. There were 2 CGS-WG sessions and 1 joint session with DAMED/NM/CGS.
The latter session was to figure out if/how these groups should interact. The
DAMED group is shutting down (finished) and the NM group will work on a schema
that might later be fed to CGS for translation to CIM.
The first CGS-WG session reviewed the group charter and goals, and a brief intro
to CIM Modeling was presented by Ellen Stokes of IBM. The second CGS-WG session
reviewed the JSIM document, specifically the UML diagram. Various issues were
raised that will be discussed on the mailing list.
During today's call, we again discussed the issue of producing an XML
representation of the MOF we deliver. Peter and Jenny suggested that it was a
"contract" in our charter. Andreas suggested we wait for and make use of work
that IBM is doing to algorithmically transform MOF into XSD. Larry said that
even with an algorithm to do the translation, producing an XSD that is relevant
and useful to the OGSA efforts will be a lot of work. This issue will be
discussed on the mail list.
Peter asked how decisions are made in a WG. Jenny said that any decisions made
during meetings or calls should be reiterated to the mail list. We should
establish a policy that decisions sent to the mail list have a 2 week deadline
for responses. If no significant objections are raised during this period, then
the decision will stand. (Of course it can always be changed later on).
The issue of job dependencies/workflow was raised during the second CSG-WG
session at GGF6. During this call, the sense was that we don't want to dive into
this problem as part of this WG right now. This issue will be raised to the mail
list.
The JSIM UML shows the BatchQueue attached to the OperatingSystem. In grid
environments, the job queue belongs to a job scheduler/execution system, not
directly to the OS. There can be multiple schedulers, and one scheduler can have
multiple queues. Andreas, Viktor and Ellen will work on an update to the UML
that will be posted to the mail list in advance of the next call.
There was some discussion about the "clarity" of the UML. Andreas advocated use
cases. Peter asked that the UML be made as clear as possible. Andreas stated
that UML by itself is not a complete story.
The next CGS-WG call will be in two weeks (5 NOV). The agenda will roughly be to
review discussions and decisions made on the mail list over the next two weeks,
and to review the updated UML that will be posted.
Peter proposed that Larry become the official CGS-WG Secretary, and there was no
objection. Peter will update the charter accordingly.