The session quickly covered the Spec with very few minor
changes and/or comments.
Then we discussed the future of BES. The following are the
notes that were taken about the discussion.
* Andrew Grimshaw raised the
need of an interop fest. GridSAM (William Lee) and Mark Morgan (UoV) agreed to
provide public endpoints. Mark Morgan has 'ogsa_run' .NET client for testing.
GridSAM will provide an all java client.
* Steven N. mentioned Chris
Smith will bring up a query on the need of a 'Holding' state in the overall
status. This will be raised in the public comments.
* Steven N.: The states in
BES are aligned with those in CIM. SAGA is looking at the BES state diagram.
* William L. & Andreas
S.: Can the client request any state changes? Yes. If the state change cannot
be satisfied or illegal, the appropriate negative response would be returned.
* Activity interface is out
of scope. A BES profile document will mandate the JSDL application type to be
allowed and the activity interface to be made available.
* Steve L. commented that WS-I
rendering can use WSRF BaseFault. Darren P. replied that for political reason
the WSI rendering should be free from any WSRF artifacts.
* Mark M. suggested we
should have the ability to express advance reservation, scheduling constraints,
utilization.
* William L. suggested we
should have the ability to express the JSDL optionalities a BES service
supports. For example, POSIXApplication, file staging bindings.
* NAREGI is using BES with
extension in JSDL to handle advance reservation.
* Stephen P. raised the
question of whether it makes sense to expose resource utilization in the
information model. Donal expresses concern in whether this is in scope for a
'BASIC' execution service. Stephen P. thinks this might raise false
expectation.
* Steve McGough suggested
there's a need for querying job based on attributes.
* Ali A. commented whether
the generic state model can support other JSDL application types (e.g. Web
Services). Can we abstract it further?
Darren Pulsipher
|
|||||
|
|||||