OGSA-WG teleconference minutes ---- August 21, 2003 =================================================== * Participants Takuya Araki (NEC) Dave Berry (UK NeSC) Bill Horn (IBM) Fred Maciel (Hitachi) Takuya Mori (NEC) Frank Siebenlist (ANL) Chris Smith (Platform) Dave Snelling (Fujitsu) Ravi Subramaniam (Intel) Jeffrin Von Reich (HP) Ming Xu (Platform) Andreas Savva (Fujitsu) -- (Minutes) Hiro Kishimoto (Fujitsu) -- (Chair) * Approved revised minutes of July interim. * Approved minutes of Aug 7 teleconference. * Approved minutes of Aug 14 teleconference. - (Dave) Did the session proposal drawn up in this teleconf reach Hiro? - (Hiro) Got a revised version from Jeffrin. * Agenda bashing ** Upcoming F2F preparation *** If you are attending please let Hiro know asap. (For logistics; no special permission is needed.) [Yes] Fred, Dave, Takuya(Mori), Frank, Hiro, Andreas,... [Maybe] Bill (clash, but will try), Ming [No] DaveB (clashes with UK all hands meeting), Chris, Takuya(Araki) *** Topics for F2F **** CMM-WSDM - (Dave) Already asked Hiro to allow time for a CMM-WSDM discussion. Give people an update of what's going on and how it affects the ogsa-wg. Not clear if any WSDM people might join the F2F. **** Factoring - (Dave) Look at CMM interfaces, data factorization (Steve/Ian/Jay proposal), etc. - (Ravi) Also fabric. How things come together. The communication part. - (Dave) One aspect is portType inheritance (extends). The other how things fit together (uses), including orchestration, serviceDomain, etc. *** Platform document review - (Hiro) The platform document is the main focus. We should prepare reviewable text for most of major sections before F2F meeting and review it. - (Ravi, Dave) Don't want to get bogged down on wordsmithing. Detailed checking can be down in teleconfs/email. # Agreed that text review must also be done but strict time # limits must be set (and kept!). *** Preparation for GGF9 (Ming) We should also discuss the strategy on approaching WGs. Ming is already talking with/participating in GRAAP-WG. GRAAP-WG is familiar with the platform text on Program Execution that Ming submitted. Desire to standardize job terms in JSDL. Steve (Tuecke) is evangelizing to group chairs. # Ming to check whether anyone from GRAAP-WG will be joining the # F2F. # Hiro to send out draft agenda for F2F meeting before next # tele-conference. * GGF9 sessions (Ming): What's supposed to happen with each group? (Hiro, Fred, Dave): Talk how they fit in OGSA, get input, commitment. # Agreed GGF9 target: # - Check what each WG has in mind and whether it matches the # architecture the OGSA-WG is working on. # - If there is a match, great. # - Otherwise try to work through differences and decide who # changes # - Finally get commitment from each group on a common direction. Things may not look that cohesive right now. OGSA-WG should end up providing more coordination. Other groups must also want it. (GGF is not a top-down organization.) - (Ravi) Focus on defining functional aspects; bring in more people/groups to get as much coverage of scenarios as possible (for validation). P2P was raised as an example. - (Andreas) (OGSA) P2P-WG is interested. They have already submitted their use cases to OGSA-WG. - (Hiro) P2P is not on our session table. # Hiro to do another revision of the proposed sessions and send to # Stacy. # Volunteers confirmed to lead the following sessions: # - Program execution (#3): Ming # - Infrastructure (#4): Dave # - Data (#2): Ian is our choice (given the recent data services paper) # # Ideally pick 2 people per session because there may be last minute # conflicts. * Platform document ** Minor update Andreas(delegated from Hiro) did a minor update of section 5 to fix structure (remove duplicate sections) and put in Ming's contribution on Program Execution. It is in GridForge. - (Ravi) We also agreed on a new outline, e.g., changing Chapter 2 to remove use case discussion. - (Andreas) It was agreed on the teleconf 2 weeks ago, but wasn't approved in the last call. In particular we didn't check the following revised proposal from Hiro: http://www-unix.gridforum.org/mail_archive/ogsa-wg/msg00338.html # Agreement that structure in email looks ok, but we are not quite # ready to approve it right now. Deferred for the F2F. - (Ravi) Need to add "service composition" to Chapter 4. # Hiro to add "service composition" under Chapter 4 in the document # outline. # Agreed not to switch sections 4 & 5 order. ** Program Execution section (Ming) - (Ming) Will add a queueing service. Would like feedback on - Other missing stuff - Level of detail - Are we stepping on anyone's toes? (Were the correct groups identified?) - (Ravi) More factorization can be done in this area. - (Hiro) Reservation service: keep job under control of queueing - (Hiro) i/f of reservation: What is manageability interface? - (Ming) The serviceData of the reservation service shows number of reservations. Each reservation is a grid service exposing its details via serviceData. - (Ravi) A scheduler is not necessarily needed. Some applications might not require it. (Globus was raised as an example here. The user can target a resource (GRAM) directly.) We have to say what things may be optional, and possible compositions. - (Ravi) Do we want a more granular approach? E.g., bunch of services provide a function. - (Dave) Taxonomy level can be a single service; but made up of a number of services to define a high level capability. # Continue discussion on the list and at F2F. * AOB ** (Jeffrin) HP has released smartfrog to the community. Will also be doing BoFs on provisioning/deployment in the next GGF. Which area would be appropriate for such BoFs? - (Dave) Scheduling/Resource Mgmt or maybe ask Charlie (Catlett). - Some references: - gridweaver.org - http://www-uk.hpl.hp.com/smartfrog/ - or search for "smartfrog" ** gridforge - Group members can submit tracker items. - Need administrative privileges to change status or assign items. # Hiro to give admin privileges to Jeffrin (at least for use case # tracker). - Site suggestions can be posted and gridforge admins will look at them.