OGSA Telecon Minutes -- 11 April 2005 Attendees --------- Mark Morgan (Minutes) Andrew Grimshaw Allan Luniewski Manuel Pereira Steven Newhouse Hiro Kishimoto Andreas Savvas Soonwook Hwang (NAREGI) Fred Maciel Minutes ------- Minutes for April 6th approved Updates on F2F -------------- No update for F2F. Steven talked to Steven McGough and there is a room available for Sunday - Wednesday at Imperial College. Agenda Bashing -------------- Additionally, Andrew to summarize his meeting with EGA in Florida EGA Meeting Report ------------------ * Andrew met with EGA technical comittee for 3 hours on Tuesday afternoon. * He has a copy of the enterprise grid alliance reference model. - He can't distributed it generally to the group, but he can give it to two or three people if important. * Very interesting discussion * Fairly wide ranging * Most imformative aspect is that they are focusing solely on the management aspects. - What does the grid look like to the sys. admin user - How can they simplify the management in that environment. - How do we take care of management and provisioning of resources over time. * Their initial scope is a single data center, not multi-disciplenary and for the most part, not geographically distributed. * Life cycle they hit on a lot * They have a lot of humans in the loop -- not all automated * Spent a lot of time on graph modelling of componentization of things. - Things are hierarchically arranged components * A lot of driving use case. - They say that our use of use case is opposite of the industry standard. * Lot of big people there * Action items that we would like to do: - They would like to have a presentation in Chicago. They want to discover which working groups within GGF and OGSA need to be briefed on what it is and where they are going with the resource model. They are also willing to do on phone, but prefer in person. - To find consumers for the use cases in the ggf. Their use cases aren't as abstract as the ones we have. - Would like to get them involved in BES -- a lot of BES is management of execution services. They said that they would do that. - The IP statement in their doc. right now. A) we can't get to them easily B) it's asymetric -- they can quote out of our docs without permission, but we can't out of their documents. - They have a glossary as well, many of which overlap, and we'll have to figure out how to get them to fit wrt one another. * They don't want to write specs. Rather they'd like to work with us to make sure that our specs match there requirements. * Suggestion that we have a "brief out session" with them at GGF 14. We'd like to allocate some time for this. - However it's scheduled, it shouldn't conflict with other ogsa sessions. - Reserving session -- Hiro will take care of. * Who will identify and invite related WG. - Andrew is gonig to send mail back to Paul Strong. * The people in OGSA who should see this doc., like Fred, and then we just need to go through list of WG and see which ones will have a lot of overlap. So, inside of schedule resource mgmt there would also be a number. Andrew taking this on as his action. * Their document will be ready in about 3 weeks. Content not going to change, but marketing will. Profile Policy Statement ------------------------ * We didn't have enough attendees at last meeting. Any one who missed it who has something to say? No one. * Steven sent comments off to Dave Snelling this afternoon. - Generally was just suggestions for re-wording and clarifications * Dave and Hiro to report back to GFSG call tomorrow. WG Status Reports ----------------- ByteIO (As per slides) OGSA-D (As per slides) - Charter likely to need one more week. - Not really going to make a huge difference to what they are doing. - Dave wants to move OGSA-D F2F to later in week. OGSA-BES (As per slides) - Basic Profile Question? There are some people who don't want to use OGSA if you don't use the Basic Profile. At the moment the Basic Profile is WSRF based. If your working group or the product of your wg has the name OGSA in it, then it should be referrant to one or more of the basic profiles. -- Would like to punt this discussion A) to next slide, and B) until official policy statement on this resolves. -- Some people think that it shouldn't have the ogsa prefix without basic profile. Who? -- Ian Foster -- not sure about Steve Tuecke - There is already a mailing list for BES. Andrew's been using OGSA mailing list but there is ogsa-bes-bof mailing list. - WSRF conformant doesn't mean that you have to use all WSRF functionallity. -- True, but it doesn't say you can use other funtionallity. - Should the OGSA basic profile be used even if the OGSA elements aren't excersized. - But you are restricted from using alternative notification. -- Some people are very uncomfortable with that. -- This is why there is desire to defer this until after the steering group policy is out. -- Seems like this might be a matter for the WG to discuss. - So, should the question be, "Will you be referant to 'an' OGSA basic profile", and not 'the' OGSA basic profile. -- If you don't use notification, resource properties, or lifetime, then you can be basic profile conformant. -- However, if you use an alternative for those very things, then you aren't profile conformant. - This discussion seems slightly out of sink with other discussions right now. - You can still have a WG without OGSA basic profile but don't call the wg ogsa. -- Yes, but he can if he writes another OGSA basic profile. - Andrew believes that we can put together a set of interfaces/service descriptions that identify the mechanisms and pieces we want -- several of which have multilpe choices. - One way to punt around the problem is that if you scope it narrowly enough, you don't need notification, resource properties, etc. But then, you eventually end up with something with no decisions. Would rather really find out what is needed. - Is BES concerned only with Jobs, or also with Queues. -- Well, a BES container may encapsulate a queue, but really it will be about what meta data is there to discover about the queue and how it works, and also about the job state. Not a queue itself though and definitely not a meta-scheduler. - To be clear, this is only a piece. This assumes that a decision about where to run has already been made. - Is this API just for the container, or also for other components. -- Two scope issues -- legacy, and service instantiation. -- Two interfaces -- the container, and the interface of the handle for the job that you get back. - Important to distinguish betweem APIs and service interfaces. OGSA-naming (As per slides) - GFS has significant concern and interest in compatibility of RNS and naming group. -- They believe that RNS is a significant work that should be used and should be seriously considered in the charter and in the charter description. -- But is it part of the charter to continue to refine and deal with comments as we go forward in this wg, or should GFS be doing it. -- The basic plan is that this is going to hold the same role as the renewable reference stuff. The advantage to this is that it is even lower threshold. - Don't want to use the word substitute when talking about renewable reference. We're trying to keep away from the problems that occurred with rr. - Andrew should create a new tracker for wg approval. - RNS inclusion seems to have been resolved now -- it is included. - What about XRI? - What about Handle System? -- Need to look into those things. - About RNS specification. It isn't already submitted to GGF editor...when will it be submitted to GGF Editor -- before GGF 14. Wrap Up ________ * Wed. is roadmap discussion. - Hiro to write up something to send out. * ogsa-bes-bof mailing list has been approved. - When you create a wg, a new mailing list is created.