OGSA Interim Meeting at OGF20 ============================= - Venue: University of Manchester - Date: 11 May - Agenda details: https://forge.gridforum.org/sf/go/wiki1714 - Minutes: Minute taker is noted per topic - Participants: Participant changes, if any, are noted per section * Participants Hiro Kishimoto Jem Treadwell Andreas Savva Donal Fellows Ellen Stokes Dave Snelling Hugo Mills Michel Drescher Steven Newhouse Ravi Subramaniam Steve McGough Chris Kantarjiev ZackK Fred Maciel Ferenc Szalai Adrian Toth Fred Brisard Bahareh Rahmanzadeh Abdeslem Djaoui Sergio Andreozzi Laurence Field Bridge: Frank Siebenlist * Summary of Actions AI-0511a: Ferenc Szalai to confirm NorduGrid/KnowARC participation and potential co-leader of IS design team. AI-0511b: Michel to pulla workflow strawman poposal together AI-0511c: Steve Newhouse volunteered to look into the possible next documents for the EMS Architecture and draw up a proposal AI-0511d: Hiro to do another zoomerang to confirm summer F2F dates * Glossary review [Minutes: Andreas Savva] - It is still an open issue whether to publish this version of the Glossary as new GFD or not. - So far this document has been synchronized with the OGSA Architecture document, but work on that document has not started yet. - There is a possibility to do an errata release for the OGSA Architecture for just the Data-related definitions. ** Discussion of terms [Some of the discussion captured below. The final text is in the draft edited by Jem during the meeting.] *** State Following last call and list discussions - Should text on 'stable' (non-changing) state be removed because it is 'not interesting'? - Is state a 'snapshot' of interesting values? No, state is not a snapshot; but a snapshot can be taken of it. - Deleted second paragraph (new paragraph introduced after list discussion) - Clarified what 'change in state' means: it is a change in the value of an attribute - Do not discuss configuration, or define static or dynamic state; might use informally *** Abstraction - Discussed whether to define this term or not, or say it is dictionary definition. Agreed to remove *** Virtualization Reworked definition at length. See document. *** Data definitions Skipped revised definitions. These should be relatively uncontroversial and are result of discussions within the OGSA Data WG. *** Reviewed status of other terms Discussed and agreed to delete the following terms: - Grid component: eventually needed but a more precise definition will be provided by the RM-WG - {Commercial, Technical,} Enterprise applications - Enterprise grid - VMM - acyclic digraph - Grid computing - ... Agreed to discuss further on a call: - Network element - Pool * Security - AOGS-AuthZ [Minutes: Michel Drescher] - Frank Siebenlist summarises AuthZ sessions at OGF - Discussions are reported around caching session management - Collaboration progress report by Frank -- Authorisation call out interface -- Concept of obligations (identifiers to call-out handlers) -- Obligations need to be standardises for the PDP [PEP?] needs to know them -- complete certificate representation in the rtequest context -- for this, the attribute names under which those certificates are registered need to be standardised, too - Hiro: Caching session management, is there another standardisation group like OASIS that takes care if this? -- Frank: Not Grid specific indeed, but there is no group in OASIS known that standardises similar technology such as XCAML - Hiro: This is very important and basic issues, but why does the WS-Security group does not care about this? -- Frank: [...] - Hiro: AuthZ discusses an architecture document, is there a roadmap for such document? -- Frank: The group thinks there is still more discussion required before going into the details of this document - D. Snelling: What's the timescale for the plans? -- Frank: I don't know. THe standardisation of the attributes mentioned above are foreseen to be finished in the next couple of months. The CVS document is much more complicated document which will require considerately more time. - D. Snelling: There is nothing more recent in the AuthZ group other than GGF14. Can you give us a pointer to more recent versions? -- Frank: Indeed - there is also no documents referenced in the meeting info for OGF20. I will send out the relevant documents that were posted on the AuthZ mailing list to the OGSA mailing list. * OGSA EMS Scenarios [Minutes: Andreas Savva] - The document has finished public comment. It has been updated and also completed the a final call on the list - Agreed that there is no problem with resubmitting the document to the OGF Editor for publication * OGSA Information Model and Information Services [Minutes: Andreas Savva] - Confirmed that Laurence Field has permission to lead the Information Services design team - NorduGrid/KnowARC will look into providing more participation and perhaps put a name forward as a co-leaded of this team - The overall plan moving forward is to arrange teleconference slots on OGSA calls; set up the design team and possibly propose a new WG at the next OGF AI-0511a: Ferenc Szalai to confirm NorduGrid/KnowARC participation and potential co-leader of IS design team. ** Adrian Toth presentation on Requirements (See uploaded separate notes) - Aggregate all information, or select which information should be aggregated - Does IS need to understand the contents? Main concern with scalability and performance. - Is it just to deliver information? Need also to decide whether push or pull model or both - Aggregate information and then how to query it. XQuery as a candidate. - Agreed to give Ravi some time to repeat part of his presentation from the JSDL XQuery session - Lifecycle of information and problem with it always being out-of-date because of distributed nature of the grid - Logging or provenance information should be separate but probably in similar infrastructure - Uniform type of aggregation? - Doing more than collect and aggregate might require some common base aggregation types to enable aggregation without needing to understand all the content - Snapshot of the whole is not possible; but it should be possible to take partial snapshot efficiently - Discussed whether general or specialized IS is necessary and what the problems with scalability etc are. Main concern with the general approach is that it might lead to lower performance. - 'General' should really just mean: - same information model; same (or similar) interfaces; and different performance characteristics, specific to domain - Agreed that goal is not to design an information system; but to identify the interfaces, info model etc. Building an information model is a different exercise. ** Ravi's presentation on XQuery - slide 7: similar to CMDB? CMDB is a concept. This could provide same functionality. - slide 8: it should be possible to set lifetime; but lifecycle is more complicated because it depends on environment etc - slide 10: both DB and XML DB and this drives the requirement for XQuery because it can provide same functionality as SQL - data is synchronized between stores; just looks like a single db - primary form is XML; incoming data is XML and it updates relational db - relational db is for large queries and to provide better performance - Aggregate is based on what clients require - slide 13: access control? What kind of information can be 'seen'; for subscription, etc. - How many machines: 400 at the time. - Relation with OGSA RSS? Could be implemented as query front end to this kind of information service. - RSS in its current incarnation does JSDL rewriting in addition to querying - Should this be a different interface or one for answering a particular set of questions? - Perhaps decompose interface as separate query and rewrite for jsdl; rather than an interface that refines jsdl - ... ** InfoD connection - Need to check what INFO-D has done in this area because it has been designed for this kind of problem. There is a short and dense document published - If it does not fit requirements might be worth pursuing changes to INFO-D - INFO-D does not use a specific language for the filter Chris K is the primary contact for information regarding INFO-D but the IS team should be the one to look into whether INFO-D satisfies requirements * Participants Changes Leaving: - Jem Treadwell - Ferenc Szalai - Sergio Andreozzi - Laurence Field - Bridge: Frank Siebenlist Joining: - Reagan Moore - Bridge: Fabio Benedetti * Workflow [Minutes: Michel Drescher] - Chris K recaps the workflow session that were held at OGF20 - Adrian now gives a presentation about WfMC, Workflow Management Correlation (http://wfmc.org) - WfMC is a member of OASIS and OMC - XPDL and Wf-XML/ASAP standards exist - workflow design patterns, terminology and glosary are available - Workflow interop levels -- no interop -- coexistence --- no standard approach --- products share execution environment --- human interaction -- Unified Gateway -- API -- Shared definition formats --- XPDL 2.0 is proposed for use in OGSA - 4 abstract interop abstract models defined - Nested synchronous sub-process - XPDL vs BPEL -- XPDL is abstract, BPEL is concrete -- many similarities -- different, yet complementary - AI-0511b: Michel to pulla workflow strawman poposal together - Abdeslem presents a couple of slides on OGSA and workflow -- Basic aspects --- Addressing instantiation, hosting, deployment, etc -- Differential aspects --- Managing state --- Asynchronous comms --- human interaction, is it possible? - Example: BPEL --- abstract and executable processes --- a workflow is exposed as a web service --- asynch PtP comms using events - Example: Microsoft WF -- Based on XAML or .NET -- suppots basic aspects as described above -- plus, it allows direct programming as code snippets in the WF -- can be exposed as a web service -- Export/Import WS-BPEL - OGSA perspective -- worflows are services -- abstract process definition -- What's the OGSA scope? - The group enters an open discussion - Macro style extension of BPEL is not possible, such would be BPEL sub-processes - If using BPEL would restrict and make impossible certain features OGSA needs - OGSA already took the decision that workflow interop is out of scope - The group discusses whether OGSA should endorse BPEL and move on, or rather keep silend on workflow and orchestration - XPDL at first pass seem less appropriate - A big issue is the integration of workflows with the current/evolving security infrastructure - Probably best to write a primer/best practices on integrating BPEL with OGSA? - We have a need to compose several BES instances together - More important, composing BES instances and DMI instances together in order to meet the full JSL requirements (stage in, execute, stage out) - Executing an application can actually be observed as a workflow all by itself (provision, deploy, configure, instantiate, execute, etc.) - The group would like to collect a list of scenarios that describe workflows in greater detail - The group reached consensus to create a Community Practice track document summarizing the experiences taken so far with workflows. * Participant Changes Leaving: - Abdeslem Djaoui - Hugo Mills * General session [Minutes: Andreas Savva] ** Document status updates Updates done by Hiro directly to the status presentation. Brief notes below only on major issues. *** EMS OGSA EMS Scenarios was submitted to the OGF around noon and has been published in record time as GFD.106. There are two other candidate documents for the EMS design team but there are no firm dates associated with them and no work started. AI-0511c: Steve Newhouse volunteered to look into the possible next documents for the EMS Architecture and draw up a proposal *** Interoperability The interoperability session during OGF20 raised a number of questions about the commitment of the parties. There seems to be backtracking from the initial agreement. There is also another cloud due to the new profile proposals. Other specifications mentioned in that session are in the process of doing their own interoperability and do not require an OGSA-led effort. There is no consensus that this interoperability testing is needed. [Participant Changes: Leaving: Fred Brisard] ** Tasks Updates to the tasks by Hiro directly to the presentation slides ** F2F discussion According to the (partial) results of the zoomerang survey the best dates are Aug. 13-15. Some late submissions to the zoomerang (Andreas) were not included, however. Discussed and produced a tentative F2F agenda to be uploaded by Hiro to gridforge. AI-0511d: Hiro to do another zoomerang to confirm summer F2F dates ** Teleconference schedule - Discussed the time slots and various proposals for different times and decided to leave as is - Discussed topics and updated in line; hiro will upload and need everyone to confirm - Agreed that having a logistics slot every week is too much. Agreed to put three logistics slots every 4 weeks; leaving one week between a Thursday and a Monday slot. ** Leadership - Discussed and drew up a short list of candidates: - Andrew Grimshaw - Chris Kantarjiev - Steven Newhouse The candidates that were present were asked to consider the time commitment involved. Hiro will contact Andrew Grimshaw.