GGF12 OGSA EMS design team session #2
=====================================

Sept. 22 (Wed), 1:30-3pm D008
note: Hiro Kishimoto

* Ravi goes over his slides. Ravi is going to list up related WG/works
starting from his list.

<CGS>
Hiro: They did modeling of queuing system but does not do
      generic scheduler.
Jay:  they are doing very good job with WS-CIM WG.

<Community Scheduler Framework>
Andrew: Chris says CSF well fit current EMS.

Jay: Within next 6 month, we should complete sure spec
     of very small set of services.
Ravi: We will collect information from all related WGs to
      find out what they will do and identify gaps, we may tackle.

Jay: List up existing work and ask them to come back
     again 6 month later is risky.
     we should draw up expected services, e.g. "job",
     and ask experts to do such and such things.

Ravi: But we don't have any information now.

Jay: "job model" has very critical problems.

Hiro: Let's table process discussion and start roll call (WS listing
      up) as our first step.
Ravi: (to audience) Please explain what are you doing.

<GSA-WG>
Ramin:  usecase and scheduling description.
	EMS should leave room for general description
	are the sub-component is flexible enough?

Andrew: current breakdown is for maximize flexibility.
	you can compose any of these when coding.

<Workflow RG>
	They work on requirement capture.

<JSDL>
Ali:  Description language for job.
      Scheduling and management is out side of scope.
      Term language and structure for the job.
      Want to understand the EMS's concepts and see how it fit to our
      work. 

<CDDLM>
Dejan: We do "deployment."
       Deployment includes 1) bring things in, 2) configuring, 
       3) Life cycle management.
       We will define a) language, b) basic service interface, c)
       component semantics.
       By next GGF, we will complete our specs.
       LM is just for deployment.
       We don't support versioning. (hope ACS does this)
       Now we have very limited deployment APIs.
       Looking forward to support lifecycle events.
       Resource allocation is out side of scope.
       Other efforts in this space; CDML, IBM Solution Installation.

<ACS>
Hiro: CDDLM is a client of ACS.
     
Ravi: Dynamic deployment is required. For that, you should record all
      modifications for backup recovery. We have such implementation
      already. 

Pate: We should draw line between application parameters and job
      parameters.
  
Ravi: Does CDDLM use job model for deployment? do you use any
      container model for you deployment?
Steve: No, but it may help.

Dejan: Do not widen CDDLM's scope now.
       This gap analysis is the same as we found at our demo.

Ravi: Each company do its own development already, but for
      interoperability how to connect them is issue.

Ravi: When you want to install OS in the box, what software
      should run on the box?
      Big picture should be combine all these pieces together.

<Policy RG>
	Darren Pulsipher (chair)

<Logging>
	Bill is creating new WG

<out side of GGF>
     UT WG - DMTF
     DCML - OASIS
     WSDM - OASIS
     CIM - DMTF
     WSRF- OASIS (basic)

<Next Step>
Bill: Focused interim F2F meeting works and helps.