[rm-wg] Meeting Minutes, 20th April 2007

[Strong, Paul]

RM-WG Meeting Minutes
20th April 2007
 
Attendees
 
Paul Strong - eBay
Dave Snelling - Fujitsu Labs Europe
Duncan Johnston-Watt - Engimatec
Chris Swan - Credit Suisse
Ali-Anjomshoaa - Fujitsu Labs Europe
 
Agenda
 
Approval of minutes - approved unanimously
Discussion of life cycle diagrams - see
https://forge.gridforum.org/sf/wiki/do/viewPage/projects.rm-wg/wiki/20th
April2007
 
Notes
We spent some considerable time discussing the life cycles as originally
defined in the EGA Reference Model v1.5 and as proposed by Dave
Snelliing.
 
The definition of Created/Extant seemed simple - the grid component is
manageable, i.e. the GME (people and/or technology) "knows" that the
component is present and the GME may manipulate or manage the component.
The proposed alternative name - Extant - was also briefly discussed.
Some argued that this was too abstract.  Deciding upon a final name was
deferred until state definitions (see Configured below) were better
defined, so as to avoid stalling progress. Dave Snelling observed that
the time interval between the act of creation and destruction mapped to
WS-LifeTime(?) 
 
The definition of Active also seemed simple - the grid component is
serving the function for which it is intended, for example a service is
responding to or is ready to respond to requests, a physical compute
device is powered up, and operating system is booted up and any
authorized user can log in and so forth.
 
Difficulties arise when we think of the Configured state.  Especially
what are the boundary conditions of the state.  Clearly any manager of
the system (GME) would want to be able to determine whether the grid
component is in a state where it can be made active (i.e. started), or
whether recovery is required following some failure whilst previously
active, or whether it can no longer be made active for some other
reason, for example there are limits on the number of times a specific
component may be made active.  Dave Snelling's diagram reflects this
desire in the form of additional state and state transitions.  Paul
Strong pushed back with the desire to maintain a simple model if
possible, suggesting sub-stating the Configured state and adding
transitions.
 
After discussion, Paul Strong suggested that the group work a number of
examples using both cycles to tease out a clear definition of the
Configured state.  This should help us determine whether additional
states should be accommodated as sub-states of Configured or as peer
states.
 
Actions
 
Members to work examples for both the orginal and the extended state
transition diagram.  Please send to rm-wg at ogf.org to promote discussion.
 
EFI - Dave Snelling
BES/Job - Paul Strong
Server - Duncan Johnston-Watt
OS - Paul Strong
Database Instance - ?
Service Component (EAR/WAR) - ?
Web Tier (inc. Load Balancer and multiple Web Servers) - ?
FileSystem - ?
Autonomous Service - Ali-Anjomshoaa - Fujitsu Labs Europe
 
For those who were not present at the meeting, please feel free to
volunteer and take ownership of one of the unassigned above.
 
Next meeting is on 27th April 2007 at 8AM PST (see
https://forge.gridforum.org/sf/go/projects.rm-wg/wiki)
 
 
-------------- next part --------------
An HTML attachment was scrubbed...
URL: http://www.ogf.org/pipermail/rm-wg/attachments/20070425/8e4c317e/attachment.html