OGSA Teleconference - 26 October 2005 ===================================== * Participants Mike Behrens (R2AD, LLC) Chris Jordan (SDSC) Andrew Grimshaw (UVa) Hiro Kishimoto (Fujitsu) Allen Luniewski (IBM) Steven Newhouse (OMII) Kazushige Saga (NAREGI) Andreas Savva (Fujitsu) Ellen Stokes (IBM) Ravi Subramaniam (Intel) Jem Treadwell (HP) Jay Unger (IBM) Pete Ziu (Northrop Grumman) * Oct 24 minutes approved with no changes * Information model discussion - Ellen posted an initial draft of a mapping of GLUE to CIM to the list before the call. - This is still ongoing work; material is incomplete. - Ellen has also had the opportunity to talk with some of the GLUE authors this week and has gained a better understanding of the model. In particular, some elements in GLUE have had no experience; they may be included as placeholders to allow for extensibility later on. - Explanation of the MDS-GLUE-BES model (pdf file) - Items in blue are required by GLUE - Items in black are already in CIM - Items in red: - if {D} they deprecated by CIM - if {E} (experimental) they are new and need experience before complete acceptance - 'Blue' is GLUE-required and in CIM (or equivalent) - There are also a number of GLUE items that have not been mapped yet and so are not in the diagram (these are listed in the Excel sheet) - One example is 'site' - It is not in CIM and seems reasonable to add; but haven't done so yet. - Items such as 'Load' may map to CIM metrics instead - Some items may map to a combination of existing CIM entries: e.g., 'StorageDevice::Size' - For items such as ClockSpeed (l.162), and MTU (l.157): - These are defined as having one value while CIM has equivalent items that have two values: Maximum and Current. - The meaning is probably equivalent to 'Current' but need to confirm. - Also there are some items, e.g., lines 165-168, in the Excel sheet, that are not clear how they will be used. They have been added to GLUE on the assumption that they might be useful but there is no experience with them yet. - It was discussed how such items should be tackled. For example identifying the size of cache might be better abstracted as a memory performance metric. - Why is vendor model version needed? It seems like a composite identification for particular chip and chipset and might be required for specific codes. (Scientific codes may have different requirements.) - Following the above discussion it was clarified that the purpose of this modelling work is - For things that are obviously missing from CIM to extend CIM to include them; and - For things that may be too domain specific(*) and a CIM extension may not be appropriate at this time to show how they can be included without extending CIM formally. There will be no standard mapping for them but they can still be used. The guidelines document will show how these can be added to CIM in this way. [(*) These were called "situational extensions" during the teleconf call] - All these 'situational extensions' will be listed in the final version and will not be defined as either CIM or OGSA 'extensions' ** Slide 3: BES preliminary Action: Andrew to review and point out what is needed for the BES container - Goal for next Wed: A preliminary set of attributes that are needed for the BES container - Additional attributes (Network, Cluster, etc) to be added later incrementally. * UML tool selection - Probable usage is sequence diagrams and information modelling - Prefer a commercial tool (because it may be better supported) - Easy to use is a big advantage - RSM is commercial, has rich functionality but there are some concerns on how easy it is to start using it. - Also whether the IBM Scholar program is available only to US and Canada. - Agreed that it is not possible to make a decision on this call. - One of the IBM participants will follow up and confirm the procedures needed and that everyone can use the RSM tool, if the group decides to go with it. - Mike Behrens has some experience with a number of the listed tools. He volunteered to check out the rest and give an evaluation next week - export/import as one required feature * MDC presentation review - Ravi sent out modified presentation based on Andrew's presentation that was uploaded to Gridforge. - A 45 minutes slot in total is allocated for the presentation. There are too many slides as things stand. - Not sure if it includes Q&A; someone (Hiro?) has to confirm. - Audience? It is not clear. Assume people with various backgrounds. - What should the focus be? - On technical work; on process; or motivation/goal? - It is an introductory presentation. Jay and Hiro suggested a higher level of abstraction. It was agreed to do so and also add some slides on the connection between DMTF and OGSA. - Assume they know what a grid is, or just put one slide on what grid is - OGSA: what it is, what is based on, where it is going (a max of 3 slides) - Relation to other activities - More functional description of components in OGSA; component diagrams as in EMS for other capabilities if possible (not quite ready for this yet). - Slide 26 is good as standards landscape - Slide 24 is also good; add CIM and WSDM (and transitions to call out the standards bodies) - Slide 38 is good but maybe too detailed for this presentation - Slide 16 is ok; but it is not good to explain things from this slide - slide 22; Replace with Fred's version since it includes the CIM profile (Hiro will send to Ravi.) - Are the Acknowledments correct? - Ellen will help with DMTF-OGSA slide. Again start from Fred's slides at GGF15. - Ravi will update, and circulate to collect comments. - And arrange to have a short discussion on a subsequent call - Need to send agenda and title of this presentation to MDC: - Title: " Open Grid Services Architecture" and - Subtitle: "Service Oriented Architecture for Grids" - Agenda should be a paragraph of text. Ravi to write up. * EMS scenario review. Postponed. * Next call - Agreed to cancel the Monday call since it looks unlikely to reach quorum. - Next Wed: - Add to the agenda a reviewa the state of the naming discussion/decisions.