Dan Gunter gave general overview of group history and purpose, and
outline of session.

Ellen Stokes gave a general introduction to DMTF CIM
CIM provides:
  - modeling language and schema
  - management schema
  - protocol to encapsulate
  - compliance document

Some discussion about the CIM backend, and how the CIM schema could be
expressed in different languages.

? ? from IBM on CIM relationship with DAMED
There is a concept of subscription: An application can subscribe. The
operation returns response to subscription.  The provider can
implement this by spawning off new synchronous thread to wait for
event.  

example: If the event is CPU utilziation increases beyond N, the
thread can periodically check and create an "indication" to send back
to the client which has a listener setup.   Straightforward
subscription interface.  Could also be implemented in hardware if the
hardware is capable of generating an event call.

In CIM, there is a base class for indications that explains the
general behavior.  Subclasses can be defined that provide information
on top of the base class to support additional information.  

(from crowd) There is a set of predefined indications.  Your CIM model
has predefined indications for the lifecycle events that you are
interested in.  

There was a discussion of how CIM relates to DAMED.  Does CIM have
special qualities that make it appropriate to provide information of
interest to DAMED, or is it just another way of providing information
about a schema.

Indications tree is separate from rest of schema.

Question about timestamps.  DMTF seems to have developed its own
timestamp type.  



Dan resumed giving the overview with the charter.  Need to decide how
much of purpose of DAMED has been resolved by CIM and other groups.

Then led into Top-N document.  

Jennifer Schopf presentation on GLUE-schema.  Trying to get US HEP and
EU HEP to work together.  Each side had different schemas that were
incompatible.  One problem was that different communities take
different views. Some node-by-node, others based on abstract service
providers and paths.  Original intention was to automatically
translate between existing schemas, but this was very hard with the
different perspectives.

One goal of GLUE is to define attributes without requiring a specific
hierarchy model.  A hierarchy is provided to describe clusters, but
the atttributes are designed to be more generic.

James Magowan began discussion of Top-N document. 

First question, can we only send changed events?  Is there a very
basic set of common events we can all agree on and then represent in
whatever system and schema we want.  Thus the goal for the very basic
top-n events.  

Question about use-cases.  The top-n document was written based on
common experience and belief of what we needed for work.  Some
recollection that there was a GMA use-case document on that.  

Mention of use-cases produced by Grimshaw and Legion group for
portals.

Question of whether CIM provides all information that is required or
described by Top-N.  Clearly no need to define top-N if CIM contains
entries for all top-N events.  Discussion of motivation for originally
having top-N to make grid systems interoperable.  

James gives 30-second overview of different perspective.  CIM, etc try
to produce entire schema, DAMED is trying to focus on lowest level,
just event definitions.

Sathish Vadhiyar talk on Tools for Performance Evaluation.  Using
hardware counters.  PAPI: Performance Application Programming
Interface.  Q: Has PAPI looked at providing information in a schema to
publish information?  A: no.  PAPI is only a single-application
picture of the machine---what one application is doing.  

James resumed Top-N document

How is disk information stored?  Aggregate?  Device?  Filesystem?

Profiling information interesting.  Runtime of application (both time
and other resource consumption information) of common interest.