OGSA Teleconference - 8 August 2005 =================================== * Participants Hiro Kishimoto Mark Morgan Ellen Stokes Fred Maciel Jem Treadwell Steven Newhouse Pete Ziu Ravi Subramaniam Takuya Mori Jay Unger Andreas Savva Minutes: Jem Treadwell * August 3 minutes approved with no changes * August F2F Update: - Hiro is waiting for detailed agendas from most working groups - Hope to get agendas for the next meeting * Provisioning Discussion - CDDLM WG provided a proposed agenda for the f2f - Ravi asked: Do we require some update for the Architecture document that will involve CDDLM? Will they make a contribution to OGSA 1.5, or will we just reference them where needed? Hiro: It is more likely to be in the Architecture 2.0 document, not 1.5. But we can discuss this at the f2f. * Resource Modeling Discussion - Fred said that we need to understand the next steps: need to make a decision on CIM before going too far. There followed a long discussion with the following key points: - CIM is widely used and adopted in the industry, so if we decide that it's not all right then we have to come up with our own model, which would result in all kinds of adoption problems. - If we determine that it's not the correct model, we should take the approach of making it all right - we have the support of many people directly involved with CIM. - There might be some perception that CIM is more suited to physical resources than logical - Ellen pointed out that it addresses many physical resources and some logical resources, but provides hooks for other logical resources. We should take the approach of extending if we need to. - Steven said that we're talking about resource models, but CIM would be better called an information model; agreement that the terms are being used pretty much interchangeably, which may or may not be appropriate. - Steven asked why we're focusing on CIM, and not the GLUE schema; Fred responded that we've held education sessions and discussions, and that Ellen worked on a mapping of CIM/GLUE - now looking at it again since GLUE has evolved. Jay: We'll end up looking at ways to bring the models together. Fred: would love to have GLUE team members here, but they don't have the time. - Steven pointed out that if we want OGSA principles to be adopted then we must relate them to things that are being done today. Fred said that he has asked Jenny Schopf for a liaison, but nobody in the group can represent them. Steven agreed that it's an ad-hoc group that does things as needed. - Hiro raised a concern that CIM may be too big and complex, and not focused enough on areas of importance - may not be able to successfully subset it and influence its direction. Jay said that this bears investigation - we need to discover if we can have a consistent approach that will give us a containable view of the larger model - can we then use those pieces in a consistent way, or will it be so fragmented that it's not useful? - Jay said that in arguing for CIM he is not intending to short-circuit work to be done, but would rather spend time on a concrete experiment against a framework that has broader adoption - nobody adopts all of it! People are successfully subsetting the model and we can learn from them and do a concrete experiment. If we based it on something else or built from scratch we would have difficulty, and if we were successful we might have difficulty getting it adopted. Would rather work on CIM and fail and know why we failed. - Hiro noted that Fred and Andrea already started work, but haven't started writing it up. Fred reiterated the need to discuss the order of things to move ahead - we need a broader discussion so that everyone in the WG can be comfortable with CIM. - Steven: I can look at GLUE and grasp it in 5 or 10 minutes, but can't do that with CIM. Jay emphasized that it's important to demonstrate that we can do the subsetting of the framework and the model necessary to represent (for example) BES and batch jobs. Ravi said that we need to separate resource and information models - need to recognize difference and translate between them. Steven agreed. - Pete Ziu asked if the right approach is to create an OGSA CIM profile. Fred: Yes, but we need extensions which we could submit to DMTF. We only need a small set of classes to use it in OGSA - don't need to go to the whole model. Although we have started to work on this twice, we have been off direction. However, we do have a preliminary document, and work is progressing. - Hiro said he believes Fred's strategy is right, but we need a case study to write up the document. Ellen said we need an agenda for the RM team at the f2f, so that we can go through a case study; we can use Fred's doc as a starting point. Steven agreed that he can represent GLUE in a f2f discussion - Fred said that we need to add information about GLUE to the document, but he needs somebody knowledgeable to provide the info. - Ravi said that we should also capture concerns - must answer and address those. Jay agreed, and suggested that we come up with a general notion of subsetting model and framework without fragmenting them to the point that they are not usable. We need to do the experiment, and we need a consensus on the conditions - may fail, but also need to establish the criteria for the result. There was agreement that we can start from the work that Fred and Andrea did. - Steven clarified that we need something that is simple and that anyone can understand - if that happens to be a subset of CIM that's fine. - Fred asked Steven to provide input to the RM document on what he would like to see; Steven agreed to do that. - Fred said that he's looking for help on what to do for the RM discussion at the f2f - What do people want? A tutorial? Jay asked for something to look at in advance - we need a scenario-driven agenda, and then a concrete discussion about what's in CIM. Fred will provide an agenda after Tuesday's RM design team call.