OGSA F2F #15 - Day 1 afternoon - BES (1)
========================================
* Participants
Chris Smith
Dave Snelling
Michel Drescher
Andreas Savva
Ravi Subramaniam
Steve McGough
Pete Ziu
Glenn Wasson
Marvin Theimer
Hiro Kishimoto
Nick Werstiuk
Ellen Stokes
Jay Unger
Andrew Grimshaw
Bridge:
Darren Pulsipher
Peter Lane
* Summary of Actions
ACTION: BES-WG chairs to ask the data group to define the
filesystems as their work product
* BES State Model
- Rough agreement to extensible model: base (3) and extensions
- ISSUE: unidirectional only; there is (at least) one system that
can put running-pending. (The point here is the job id stays the
same)
- Substates: how to define
- Two proposals
- nested
- need to define nesting structure;
- is this more complicated when multiple people are writing
the different substate refinements?
- nesting defined as the way the 'profiles' are composed
- profile-writer has to understand the relation of all
profiles
- parallel
- defined piecemeal; no global definition
- not fully parallel; (sub-)definitions in a single profile
may have hierarchy
- implementor will have to understand the relation of all
profiles that will be implemented
- (implementors will have to deal with different combinations in
both cases)
- PROPOSAL: known base states (pending, running, done), well known
sub-states (hold, stage-in, stage-out, suspend,...); should be able
to build the cross product of these states/sub-states
- Do it for both approaches tomorrow to help make a choice
* Trackers
[Trackers updated on- line; informational notes only]
- 1821: fixed in document; closed
- Added new tracker for "specification of container properties"
- review all definitions
- [artf5486]
- '..' behaviour is jsdl or bes issue or both?
- Not a jsdl spec issue; bes spec has to state that it is
undefined
- ROOT filesystem: BES to state that write is undefined, read
access expected but not guaranteed
- Filesystem definitions
- Need clearer definition of expected semantics of each
filesystem (working, home, fast tmp, large).
- Agreed not possible by this group to do universal definition
- Clearly state expected semantics but that there are no
universally consistent semantics for the ones above in terms
of sharing, consistency, persistency.
ACTION: BES-WG chairs to ask the data group to define the
filesystems as their work product
- [artf5485] existing mechanism (defined operation) to get
properties document which will include information on what is
supported (closed)
- [artf5512] keep RMF (closed)
- [artf5517] resolved not document
- [artf5483] open; depends when these interfaces can be ready
- [artf5488] resolved;