OGSA Session 12
Reporting session
June 30, 2005 11am-12:30pm

Note: Jem Treadwell
 

Hiro summarized outreach and Fred’s resource management sessions with slides

 

Abdeslem Djaoui:  Session #11 - Info services - see slides

 

Dave Snelling: Have you checked the OGSA Basic Profile for
Producer/Consumer interface?  Abdeslem: Yes; operations and semantics
are slightly different.  Do we want to abstract?  Dave: what is the
current consumer community of InfoD interfaces?  Little bit early to say
for certain, but there will be different types of info services that
have consumer interfaces, but they will be different.  Jay: using WSN?
Possible, but not guaranteed - wait to see how WSN matures and how it
lives with WSN.   Jay: can you say what the nature of the supersetting
is?  WSN is pub/sub model, InfoD uses pub/sub, but it also allows more
control of publishers - WSN starts from messages.  InfoD gives more
control - e.g. subscribers can determine what gets published.

 

Dave Berry: Data Arch WG - see slides

 

In Transfer, did you talk about mechanisms etc. for transfer?  Dave B:
not so much about mechanisms.  Jay: relationship with JSDL
stage-in/stage-out, and replication.  Is this single point in time
request, or enduring?  Single point in time.  Dave S: Things like
caching will build on top of this. Jay: Important to connect this to
replication & s-in/out.  Dave B agrees.  Jay: One possibility is to go
to JSDL and say there's going to be another way to do this; may need to
bump out of JSDL.  Dave B: yes, not sure if it's that simple - still
have to deal with failure.  Needs to be addressed; may be able to speed
up with discussions.  Jay: This is more change mgmt than event-oriented.
Dave B: As a result of these sessions we can write something more
coherent - need to schedule a call and get people into discussions.

 

Back to slides...

Transactions:  Pixie dust solution: How urgent is transaction-level
capability?  If important then we need to move on it... Dave B: For
commercial data grids I would think it would be essential.  

 

Steven Newhouse - BES - slides

Primary discussion today was on job state - JSDL s/in/out model is quite
complex ...  Jay: so independent pieces of stage-in/out are not
referencable? Separate elements?  Steve: right; discussed ideas.  Minor
tweaks - e.g. tell container to stop accepting new activities, and to
terminate.  Will get a new iteration of document within next week,
resume calls in 2 weeks.

 

Dave B: How is checkpointing done? No plans to put it into the container
- current model is if it supports it then it should provide an
interface.

 

RSS Charter BOF (Hiro)

Well attended, high interest, some implementations already.  Requirement
document is problematic - needs scope discussion.  This BoF will revise
the charter doc and will submit to AD soon.

 

BoF about the alternate (Minimum Web Service) basic profile... Dave
Snelling.-  slides

Steven: Recognition that serious time commitment is needed to do this
properly...

Jay: I could do a restricted WSRF operation and see the results in a
WS-Transfer - yes - lots of MUSTs so a compliant implementation could
live with a non-compliant... cool, makes OGSA a bridge, but may be
pretty hard.  Steven: yes...  Jay: possibility to seek some real
compromise in how defined.  Dave: Cannot change WSRP document ... Steven;
programming model might come out of it - how do you represent your state
so it can come out through WSRF or WS-Transfer... - Jay: is that it
would force a compliant to be a bridge - kind of neat.  Dave B: Unless
we get significant buy-in from WS-T backers (e.g. Microsoft) -
contribute resources...  Jay: there is interest from Microsoft (Marvin).
Dave: BoF mailing list has been approved, so discussion can continue on
list.

 

 

OGSA Schedule:

Resume July 6th

BP Final Call July 6th

 

Summer f2f: - discuss on mailing list

 

Consider f2f adjacent to GGF15 (Friday) - consensus that this is a good
idea, especially as GridWorld will limit the amount of time available in
GGF itself.