[rm-wg] Meeting: Fri 28th Sept, 8AM PDT, 11AM EDT, 4PM BST, 3PM GMT

[Strong, Paul]

All,


Next Meeting


When: Fri 28th September 2007, 8AM PDT, 11AM EDT, 4PM BST, 3PM GMT

Duration : 1 hour

Where: 
Con-call - The pass code is 2051278. UK:+44-800-085-6481, US: +1 888 237
8573. 
Glance - http://rmwg.glance.net, pass code 0928.

 

Agenda:

It's about time we recovered from the post summer torpor!   Dave and I
have been active in our organizations building things that use the
reference model, or at least something like it, and we have some
thoughts I think we'd like to share before we make the push for
publishing the document. 

 

Specifically from my point of view -

 

*	Whilst the notion of a pattern DAG, a DAG of managed components
and the logical DAG of the components that manage is very useful in
describing a conceptual view of how things fit together, the
sub-classing of Grid Components into Managed, Pattern and Manager Grid
Components within the UML seems to be less useful or rather less
actionable.   It may be that Pattern and Manager sub-classes are
appropriate, but it seems unnecessary to have the notion of Managed, as
all Grid Components can be managed.  Anyway, we should discuss.
	
	
*	Grid Component state.  Here we have found that we are using two
types of state, an operational state and an admin state.   These are
summary states with the complete "state" reflected by the various
attributes of the Grid Component.  I am actually uncomfortable with the
notion of Admin state because it couples the life cycle of the Grid
Component too closely to an implemented process.  It would be good to
discuss though.
	
	
*	State & Configuration.  A couple of months back there was an
interesting discussion on State vs Configuration on the OGSA alias.  At
eBay we have in fact created State and Configuration classes.  The
existing UML uses stereotypes to effectively categorize configuration
and state related attributes.  At eBay we have separate classes to
capture these.  Each Grid Component has both State and Configuration.
Our definition is that Configuration attributes are those that can be
directly externally manipulated and that define the behavior of the Grid
Component.  State attributes are typically those that are set internally
(although changes in them may be externally driven) and reflect
differences in the state of the Grid Component that are interesting to
an external observer/client (user or manager).   This State class is
separate from the state attribute that reflects where in the life-cycle
the Grid Component is.  This may or may not make sense and a discussion
would be useful.
	
	
*	Instantiation and Interaction Dependencies.    The UML is
currently incomplete because it does not allow the sub-classing of
association classes.  Association classes may be sub-classes of other
classes though.  So we would have to change the way we represent things.
Thus we would have to have Instantiation and Interaction Dependencies as
abstract super classes that we sub-class until we reach leaf classes
that can then be Association Classes.   Whilst it is certainly out of
scope at present to discuss how we could serialize the model, in one of
my projects we are using RDF and OWL DL.  RDF allows us to sub-class
classes of relationships in a much cleaner and more obvious way.  The
notion of classes of relationships and allowing these relationships to
have their own attributes is of fundamental importance.   In our CMDB
work I am pushing heavily for relationships to be fully fledged CIs.
Another thing we should discuss.

 

Other than that we need to pick up momentum again before OGF21 in
Seattle, so that we can have a document that is in a state to
review/comment upon.  I don't think we have much to do conceptually, but
we do need to get the above resolved and written down.

 

We should also get up to date on he latest GLUE work.

 

Cheers

Paul

 

Paul Strong

Distinguished Engineer

Grid Computing & Enterprise Management

eBay Research Labs

Tel: +1 408 967-8450

Cell: +1 408 981 8948

Skype/AIM: FractalPaul

http://labs.ebay.com

 

-------------- next part --------------
An HTML attachment was scrubbed...
URL: http://www.ogf.org/pipermail/rm-wg/attachments/20070927/f7a20aef/attachment.html