Cosmos Architecture Meeting 22-Aug-07

Attendees

• Dave Whiteman
• Ali Mehregani
• Don Ebright
• Mark Weitzel
• Joel Hawkins
• Sheldon Lee-Loy
• Balan Subramanian
• Martin Simmonds
• Bill Muldoon
• Paul Stratton
• Mau
• Steve Jerman
• Chris Craddock

NOTE

The meeting time/date has been moved to Thursday's at 11:00am EDT. The next meeting will be August 30, 11:00am EDT.

Discussion

Continued discussion on the design of the data broker.

• We need to have a speparate discussion involving how CMDBf affects the design.
• Need to review the attached code from Bill.

• Don:
  • We need to understand how/if the broker must understand the relationship b/t the CIs and the MDRs.
  • The current design does not support having lots of endpoints (mdw: need to request clarification)

• Martin:
  • CMDBf sits above the data broker.

• Joel:
  • We don't have instance data in the broker
  • Joel would like this, but nobody else really did

• Don:
  • There is an opaque string that may be used to help this.
  • The lack of instance data satisfies a use case of small set of endpoints, but a large number of endpoints could become unweildy

• Chris:
  • We need to stop looking at things from a "product" perspective. The key idea of the broker is to avoid falling into this siloed view. Must be able to avoid this siloed view of things.
  • Identity not in scope for COSMOS, but "logical groupings" of resources should be ok.

Joel:

• Since the CMDBf has a graph oriented result, we'd expect to be able to ask graph type questions. This means we may need more metadata in the broker.
• Patterns for finding EPRs is more pattern matching than anything else.
• This opaque string is essentially an extension point.

• Bill: Update the wiki page with use of the "classification" element.

• Mark: Need to describe the question about data manager/data context...