OGSA teleconference January 26, 2004 ==================================== * Participants Abdeslem Djaoui (RL) Bill Horn (IBM) Hiro Kishimoto (Fujitsu) Fred Maciel (Hitachi) Bryan Murray (HP) Andreas Savva (Fujitsu) Frank Siebenlist (ANL) Latha Srinivasan (HP) Ravi Subramaniam (Intel) (first 30 minutes) Jem Treadwell (HP) Jim Willits (HP) Minutes: Andreas Savva * Summary of Actions Action: Jem to add terms from section 2 of [CMM_Gap_Analysis] to the OGSA glossary once their definition is stable. Action: Fred to rework the definition of Resource Management. Action: Fred to rework the definition to make it specific to 'Grid Resource Management.' Action: Fred to check MUWS/MOWS requirements and make sure the Requirements section is consistent. Action: Fred to re-think how the 'Levels' section fits in the OGSA document. Action: Fred to come up with a better figure 1. Perhaps redoing the figure as a UML diagram. Action: OGSA-WG members to review (at the very least) section 5 of [CMM_Gap_Analysis]. Action: Fred to contact Ravi and Andrew and ask them to review document. Action: Bill to use the terms consumer/producer in the Logging document. Action: Everyone (but especially Bill, Abdeslem) to review WS-Notification and think how things fit together. Revisit this topic during the F2F. * Approved without comments the minutes from January 5, 8, 12, and 15. * F2F Agenda Hiro apologized for not drawing up an agenda proposal yet. There maybe no paperwork involved with attending the F2F and it is not too late to send Andrew/Hiro an email that you are planning to attend. Proposed topics so far: - Information Services (Abdeslem indicated that he will try to attend if this topic is included.) - WSRF (Hiro): OGSA assumes OGSI. WSRF changes things but it is too early to start thinking about adapting OGSA. Maybe do that after GGF10. For the F2F Jay Unger has indicated he would like to discuss/explain WSRF informally, perhaps over breakfast. (Also Jeff Frey will probably be attending so this is a good chance to find out more.) Agreed that scheduling an informal discussion over breakfast is a good idea. (Ravi): One issue to try and tackle is whether/when WSRF will be fully released and put in the standards process. (Hiro): For GGF10 we probably need to make some blanket statement that currently using OGSI but are thinking to refactor using WSRF. And concentrate discussion on higher levels. - Plan for GGF10 (Abdeslem): Is there going to be a new version of the OGSA doc? (Hiro): There will be one for GGF10. Not sure about the submission deadline. (Fred): It looks like there is no deadline this time. (Publish folder location when scheduling sessions.) It is not clear. (Hiro): The GGF secretary requested OGSA-WG submit its schedule early. Already asked four 90 min sessions. Sessions requested: Two on Mar 10 (one morning, one afternoon) Two on Mar 11 (one morning, one afternoon) Agreed to discuss GGF10 plans during the F2F. * Management Framework (CMM) Discussion Review of [CMM_Gap_Analysis], in particular sections 2 and 3. https://forge.gridforum.org/docman2/ViewProperties.php?group_id=107&category_id=612&document_content_id=1632 Fred posted the document about one week ago. It is still incomplete. The aim is to give a broad picture of Grid Resource Management. It is one of CMM's deliverables; an analysis to identify any gaps between WSDM and Grid requirements. The motivation is to provide a framework for management and contribute relevant sections to OGSA. One important aspect is to pin down terminology. There is a perceived overall/connection with GMA. Might want to discuss it later. ** Definitions (Section 2) *** No objections to the definitions of Management, Manager, Manageability, Manageability interfaces, Manageable entities. *** Definition of Resource (Fred) Grid and Management use 'resource' to mean different things (supply vs manageable entity). WSRF uses WSDM definitions. (Hiro): Probably need both definitions in the document. We need two terms. One idea is to coin a new term like supply-side-resource and use resource in the management sense. (Jim): Add adjective to resource to make meaning clear. E.g., 'manageable resource' when talking about management side of resource, 'grid resource' when talking about supply. Agreed to add contents of Sec.2 to OGSA glossary. Action: Jem to add terms from section 2 of [CMM_Gap_Analysis] to the OGSA glossary once their definition is stable. (Hiro): Are the definitions given in Sec.2 consistent with WSDM? (Fred): Yes, but the WSDM glossary is incomplete. (Hiro): It's ok not to have all WSDM entries as long as definitions are inconsistent. (Fred): Eventually we should also make sure we are consistent with WSRF. *** Definition of Resource management (Fred): This is the 'supply side' resource management definition. (Hiro): Referring to the list (lines 25-30) some terms/functions are not applicable, e.g., In the OGSA document (meta) 'scheduling' is a demand side activity. And 'brokering' would be on the border of supply & demand. (Jem): This list should contain control and monitoring. But is discovery part of management or something used by management? (Latha): Relationships and discovery are related but not sure if they actually fall in managemen. They are used by management. (Bill): Accounting/metering should probably not be lumped together. (Jem): Management should probably be boiled down to things like control, monitoring. (Fred): That's too restrictive. Trying to pin down the definition of Resource Management as used by the grid community. (Hiro): Maybe use a separate term then, e.g., 'Grid Resource Management.' Action: Fred to rework the definition of Resource Management. (Hiro): We agreed last time that accounting is out of scope of. Maybe delete it from the list. (Bill): Maybe leave it in as an editorial remark. *** Definition of Resource Manager Action: Fred to rework the definition to make it specific to 'Grid Resource Management.' ** Management in OGSA (Section 3) *** Requirements (Section 3.1) (Bill): Maybe narrow down the definition (l.41) to say explicitly "ogsa *resource* mgmt." (Fred): This changes the scope from Management to Resource Management. (Bill): Was looking at scope for CMM not OGSA. Suggestion withdrawn. (Hiro): Is this section consistent with MUWS/MOWS requirements? Action: Fred to check MUWS/MOWS requirements and make sure the Requirements section is consistent. *** Levels (Section 3.2) (Hiro): What is the definition of resource here? (Fred): Manageable resource (not just a grid service). From this point onwards a resource has this meaning. (Hiro): Then this section is not appropriate input to OGSA section 4.4. Section 4.4 focuses on just 'supply-side-resource.' (Fred): Ok. Action: Fred to re-think how the 'Levels' section fits in the OGSA document. (Jim): Not sure that WSDM is in the right place in Fig.1. WSDM will be used to also manage OGSA services. (Fred): The figure is not good. It needs more explanation. WSDM here means manageability i/f. OGSA services use or extend WSMD i/fs to manage resources. (Bill): Need more explanation on what the relationship between the boxes is. E.g., the relation of job submission to other services. (Fred): In OGSA resources are (WSDM) manageable entities. Resources themselves probably are wrapped (they don't implement WSDM i/fs directly.) In other words there is another mgmt i/f, e.g., SNMP/CIM, beneath. (Fred): Suggestions on how to draw a better picture? (Bill): Doing it as a UML may be a good idea. Might also be a good idea to capture relationships (e.g., composition) between resources. Action: Fred to come up with a better figure 1. Perhaps redoing the figure as a UML diagram. (Fred): Send comments on sections 4 & 5 to authors or join CMM teleconf. In particular section 5 is a walkthrough of OGSA section 6. Action: OGSA-WG members to review (at the very least) section 5 of [CMM_Gap_Analysis]. Action: Fred to contact Ravi and Andrew and ask them to review document. * GMA discussion (Fred): Initially saw GMA as a complete package. It seemed incompatible to OGSA. (Abdeslem): There was some confusion perhaps caused by similar sounding terminology that actually meant different things. E.g., the directory pattern that Bill presented is not the same as the directory entity in GMA. Another term is discovery. E.g., OGSA seems to imply that discovery is adding information to a directory. While GMA defines it as publishing information to the information system (does not explicitly goes into a directory). (GMA internally uses directory and other services. But this is implementation.) Also, a message broker is another entity sitting between producer/consumer. (Hiro): In OGSA let's think of GMA in terms of interfaces and standardize producer/consumer interfaces. (These are not higher level services.) Agreed to try and standardize the producer/consumer i/f. (Different entities may exist between them.) (Abdel): Propose using consumer/producer as more generic terms than supplier/consumer. WSRF also uses the terms producer/consumer. (Bill): Ok. They do look more standard than supplier/consumer. Action: Bill to use the terms consumer/producer in the Logging document. (Fred): Another term that maybe confuses things is event. (Bill): Often event is used when it actually means 'message.' (Fred): There was a design team proposal (by Dave Snelling, but he is not on the call) Goal: Produce text for OGSA doc by F2F. The situation is better than before, since we now understand each other's position a bit better. Two options: 1. Continue discussion on ML, or 2. Have additional teleconfs to tackle this point. Unfortunately there was no good timeslot until late next week. Agreed to continue discussion on ML. Also decided to add this topic to the F2F agenda. Also have to look at the WSRF (WS-notification) proposal and the notification functionality that is there. It is not exactly the same as OGSI. Action: Everyone (but especially Bill, Abdeslem) to review WS-Notification and think how things fit together. Revisit this topic at the F2F. ** Hiro's presentation [OGSA_and_Platform_IS] http://www-unix.gridforum.org/mail_archive/ogsa-wg/2004/01/msg00043.html *** Event (p8) Agreed that we need standard event schema (based on WSDM definition). Fred is checking with Heather Kreger on where the event work will be done. NOT an event service; just schema. Event is what is passed around. Proposal is to put the schema in producer/consumer platform section. ** Message plumbing (p9) Two points: 1. Keep message transport/binding separate from semantics. 2. We have to describe reliability (semantics) of broker. E.g., durable subscriptions. WSRF seems to be taking approach (1). Agreed that - Sending stuff from point a to b is transport and therefore comes under binding. - While durable subscriptions are higher level (i/f).