OGSA Information Services Session
=================================

 15 September 2008
 Room: Discovery

* Overview of state of work on information services by Hiro
  - Including current strawman proposal by UVA

** Questions
   - Which model is OGF pursuing? How does it fit in with other work?
     - This is just envisioned as an information service. CIM/GLUE can
       be used as the information model behind the proposed service
     - The difference is described in the document published recently
       (Information modeling in OGSA)

* Discussion
  - GLUE use case: Ask some service for the information; and how to
    get the information out of a provider; a web service interface
    that can be used to query
  - BES as the information provider has some problems; might be better
    having a simpler agent that pushes out the information required by
    an information service
    - In particular there are concerns with having such an interface
      as part of the compute service. On the other hand having a
      defined porttype doesn't mean it will be provided/run by the BES
      container
  - Information service here mainly refers to the interface, we are
    not interested in the internal implementation of how the
    information is collected, push or pull or the kind of interface
    agent/WS ...
  - Important thing is what kind of queries are needed. XQuery/XPath
    are perhaps more complex than what is typically needed? No
    consensus.
  - EGEE has specific performance requirements that may not be easily
    satisfied by an XML based solution. Is this an edge case?
  - XQuery is already used in some projects, performance is not
    necessarily an issue (for simpler deployments--simpler than EGEE)
    (David Wallom to writeup their usage)
  - Is XQuery/XPath too complex; discussion of how well it is
    supported by databases etc. No firm conclusion.
  - Don't get too sidetracked by specific implementation strategies;
    allow for multiple implementation approaches; hence the request
    for usage scenarios

** Suggested usage scenarios
  - Information about compute endpoints (BES) incl metrics to help in
    choosing resource
    - Also what applications are installed/available (Chris Smith)
  - Resource brokers (meta-schedulers) publishing into a top level
    info service and users querying---lower resources not publishing
    into this directory. [Not a use case this effort is covering]
  - How to get information out of services
    - David's document (to be made available) that describes their
      usage of glue 1.3 and 2.0 (mainly meta-scheduler?)

  - Different (query) requirements for different kinds of data?
    - Usage records..
    - Compute resources (incl. utilization)
    - ...
    - and associated performance requirements
  - Dynamic or static information is one proposed division, including
    the requirement to update it
    - what has changed since the last time it was queried? as one kind
      of query
    - lifetime associated with information? (example ldap/mds2)
    - dynamic and static info from same place? Might come from
      different places but easier to query from the same interface

* To continue analyzing usage scenarios to determine the target of
  this work.