OGSA EMS Teleconference - 23 November 2004 ========================================= * Participants Mark Morgan (UVa) Hiro Kishimoto (Fujitsu) Andrew Grimshaw (UVa) Ravi Subramaniam (Intel) Peter Ziu (NGC) Minutes: Hiro Kishimoto * Next calls Agreed to skip next to next Monday's (12/6) teleconf, since Ravi cannot make it. * Summary / Actions: (1) OGSA needs to move quicker than now. (2) Re-factoring and going-deeper is orthogonal. We should do both. AGREED: We adopt UML to specify our design. Action: Ravi: I will send another note asking their convenient time for future EMS call. Action: Hiro: Send ACS proposal to Mark and Andrew. AGREED: No another call before F2F but continue discussion. AGREED: Andrew will send out his document to small group for discussion and then send out to the whole list around say Dec. 1st.a * Early Discussion - Ratification of two previous meeting minutes: approved without comments. (Oct. 11 and 25) - Agenda bashing: Mark and Andrew sent out term redefinition draft to small group just before the call. Andrew will explain genesis of this document. * Discuss the Dec OGSA F2F. EMS topics to add on the agenda. Ravi: Have we already allocate some time to EMS at F2F? Hiro: Not yet, but I am going to allocate sufficient time to important design teams, EMS and Date. Andrew: Agreed. EMS and data should have high priority and the others are reporting status e.g. 20-30 minutes. At the last GFSG meeting (Nov. 7 & 8), GFSG are very concerned that GGF/OGSA was working more then 2 years but no implementations was come out of its pipeline. There is strong desire that some concrete standard specifications with implementations come out of the pipeline (not necessary full OGSA) within 12 to 18 month from now. As OGSA v1.1, we should focus down a couple of tractable specs to work out in short period of time. Then later put them into fully integrated picture. Charge to accelerate the process and get back the roadmap to stakeholders, otherwise we will lose momentum. EMS and data are two most likely candidates for that. Hiro: Our v1.0 document is high level and not for implementation. We need close links to existing grid projects by choosing Some services. I agree that EMS and data are best candidates. Ravi: Re-factoring across capabilities are also necessary. I am worry about duplicated efforts. Andrew: I agree. Mark and I work for multiple design teams, data and EMS, to avoid re-inventing things. Also we should realize that we cannot do everything perfectly right at the first step. We should fix them later. Ravi: There are many silos and they are doing their own ways it also makes us slow. But we need to work across capabilities. Customers are complaining that Grid software today cannot talk to each other. OGSA should not deliver the same kind of JUNK. Andrew: As an architect, I 100% agree with Ravi. All pieces should match each other. But again I think we should start from concrete and build up the system. Ravi: Re-factoring and going-deeper is orthogonal. We should do both. I object if we use ALL time for these low level things. We can use some time since there are many low hanging fruits we can take. Summery: (1) We need to move quicker than now. (2) Re-factoring and going-deeper is orthogonal. We should do both. - Design tools and UML Peter: Now we don't have formal way to describe high level Design. We should use formal method like UML. Andrew: Yes, Bill Horn proposed to use UML for our design. Ravi: Mark's document has a half-XML half-Java/c++ example, but I think UML is better for our v2.0 document. Let's use only basic aspects of the UML. Peter: There are some useful high level design tools. E.g. Together Community Edition for Eclipse http://eclipse-plugins.info/eclipse/plugin_details.jsp?id=503 AGREED: We adopt UML to specify our design. - OGSA stakeholder survey template Hiro: Dave Snelling and I am drafting OGSA stakeholder survey template. It will be available soon. We do not get enough feedback from potential stakeholders. Ravi: EGR-RG is doing similar things. EGR-RG is asking their use cases and requirements. This is complementary to yours. Two Intel guys are working on this. - EMS design team building Hiro: Though Data design team is well attended but EMS is not. Is there any way to invite people to this call? Ravi: Changing time slot may help? (Data is 10-11am EDT.) Mark: Half of data team is in UK. Andrew: For EMS, people from 3 continents should be attended. Action: Ravi: I will send another note asking their convenient time for future EMS call. Hiro: Does Ravi has any specific plan to apply WS-Agreement to EMS. Ravi: Yes. WS-agreement will enter public comment. We should check it out. - "Team redefinition" draft discussion Andrew: We've spent many times to discuss definition of task/job, etc. They are heavily overloaded and is not productive to use these words. In order to concretely describe the notions e.g. "provisioning", Mark and I decide to use new terms. Hiro: Do you plan to use these teams in the OGSA document? Andrew: No, not decided at this time. We don't ask whole Group to use these terms in the long run. Mark and I want to describe how to provision, deploy, and Configuration in general case. Ravi: How this draft deviates from the existing EMS document? Is it just re-write using new teams? Mark: This is almost the same. Only addition is "actor description". It is describing (1) How to provision the "actor" (= job). (2)Which information is necessary for provision ("actor Descriptor"). Andrew: "Actor" is somewhat equal to "job". "Solution" is sort of "Job manager". "actor descriptor" describe how to install and configure in simple case. Mark: "actor descriptor" is NOT "job document." It is abstract description of job type rather than an existing thing. Two concepts: (1) Collection of all what needed for the "job" (2) Generic way of describing "type of job" Not only job's metadata but also data themselves. Ravi: There are huge overlap with "ACS" proposal by Andreas and Keisuke. Mark and Andrew: We have not seen ACS proposal. Maybe, great minds think the same. Action: Hiro: Send ACS proposal to Mark and Andrew. Mark: Another addition to existing EMS text is detailed description of provisioning, how to caching reconcile files. The other part is almost the same as exiting text. Andreas: Propose to go through offline review for this document. Let's have another discussion at upcoming OGSA F2F meeting. AGREED: No another call before F2F but continue discussion. AGREED: Andrew will send out his document to small group for discussion and then send out to the whole list around say Dec. 1st. Meeting adjourned at one and half hour mark.