OGSA (EMS Design Team) Teleconference Minutes -- 11 October 2004 ================================================================ * Participants Larry Flon (ISI/Globus) Andrew Grimshaw (UVa) Hiro Kishimoto (Fujitsu) Mark Morgan (UVa) Hrabri Rajic (Intel) Andreas Savva (Fujitsu) Ravi Subramaniam (Intel) Minutes: Andreas Savva * Summary: - Next call in 2 weeks time; same slot - Need to start coming to consensus over terminology. A lot of terms have extra baggage so extra care is needed. (It is not an option to overload them more.) This is one area that EMS got a lot of feedback at GGF12. - Agreed on a detailed review (or at least a Q&A) of the document once everyone has a chance to read it. This review is scheduled for next meeting. - Agreed to work through Ravi's issue list. For next time focus on (b,c) and also (a) given time. - Ravi & Andrew to review V1 comments and suggest how to address them. This is work that should go in parallel to the EMS v2 work. The deadline for V1 re-submission to GGF is not that far away. - Agreed on need to refine service (interface?) definitions maybe going to the wsdl level for some. - Agreed that 'resource' needs refinement. - Need to review the material given to OGSA-WG over the summer about CIM and WSDM and try to use it in EMS work - Ravi to work on engaging more people to work on EMS. * Introduction - Document sent out by Ravi is a re-organization of the material in the OGSA spec V1 document. - In this call review goals and go quickly through this document. Decide course of action. - Reference to document updated with discussion from this call: http://www-unix.gridforum.org/mail_archive/ogsa-wg/2004/10/msg00023.html * Document walkthrough ** Focus - Question on ontology/terms: - Definition of terms and how they relate to other terms in the language. E.g., how to describe jobs. - Note that the 'terms' here do not refer just to job submission. Expect to include description of other job phases. ** Guiding principles - Minor addition: "Adhere to general ogsa design principles" (reference is ok). ** Conceptual Framework - The terminology in this area (workload, resource, etc) has a lot of baggage already - Expanded terms (hierarchy) here; includes (business) processes. - Concern raised with existing definitions and how to reach consensus. This was a major issue during the GGF12 EMS sessions. [Especially a lot of pushback on the 'job' & 'workflow' definitions.] - This is a first proposal to get the ball rolling. Expect refinement. - Need to make sure the terminology is in agreement with OGSA Glossary in V1. - The relation between these terms must also be described - The hierarchy figure is not in current document. - Ravi to consider whether to bring it back - [One problem in OGSA V1 is the relation between task and job and the inconsistent usage of these terms.] ** EMS Services/Resources - Many different types of resources; haven't talked about their granularity yet. - Past discussion indicated that the PSHS might not be a resource but some kind of information service instead. - The way forward here is to drill down on the required interfaces; that seems the only way to determine what these entities really are. - Generic properties about resources: enabled by underlying standard used; e.g., wsrf. ** EMS Services/Job management - The problem here (too) is that all these terms are overloaded - Ravi wants to put forward the concept that 'job' is 'self-aware'; in the sense that it can manage itself/enact its schedule Action: For next meeting (scheduled in 2 weeks time) need to start coming to consesus over terminology because this was a main contentious issue during GGF12. (Ravi to drive this process) ** Selection services *** Reservation - Is this a single service or set of services? - It might just be an interface. (It is somewhat misleading to talk about services at this point.) - Needs more clarification ** Interactions with rest of OGSA - Need to refine this section to the level of requirements - Agreed that it is not sufficient to just point to a WG or describe things in general. - Monitoring 'jobs' described here is one aspect. Also have to discuss the monitoring capabilities provided by EM services (what information is provided etc) ** Questions to be resolved: - Issues noted by Ravi were added at the end of the document - (d,e) The scenarios so far only dealt with what happens when jobs are started. Haven't really worked through to what happens in later stages, e.g., job migration. - (l) Delegation - Not just the security aspect but also the operational one: what functionality to delegate & how to do it. - Other issues raised in earlier discussion: - What is a resource and how to represent it - Granularity of resources - Comment that some of the questions noted by Ravi may point to specific usage scenarios. Their range can be quite big. * Discussion on how to move forward - Since many people on the call are first time participants we should allow time for them to read & digest the material and review in more detail later. - Participants should review document and be up to date when joining the teleconferences - Andrew: Need to drill down on some existing scenarios in order to determine whether what's required by EMS is provided by what other groups are definining - For example, whether deployment as defined by CDDLM-WG is what EMS requires. - EMS currently defines a small set of services. These are fairly well understood by participants (terminology question aside, in particular wrt job/job manager) so moving forward should not be too difficult. - Also resource side needs refinement - How to proceed with discussing question list - Will address as discussion topics in teleconferences (allow sufficient time so that people can think things through in advance) - Feel free to add issues to this list - Also work through some of the scenarios to get a better understanding of issues. - Given some of the scenarios we can look how the existing questions fit into them and what needs to be done to address them - Questions (b) & (c) on Ravi's list (job / resource) might be good topic for next time - and (a) as the next issue - There are a number of issues wrt v1 that have to be addressed this month for re-submission to the GGF Editor. Waiting for the next EMS call (2 weeks from now) is too late - Andreas will send the v1 issues spreadsheet to Ravi (>Done) - Address some of the topics next OGSA call / or by email - Andrew & Ravi to arrange an OGSA v1 EMS review call Friday afternoon to start discussing issues - Two proposals came up at GGF12 - Do an EMS F2F - Many people indicated that they might not want (or be able) to travel just for such a meeting - Ravi is thinking of other possibilities such as co-locating with some other bigger event. - Choose a couple of services and drill down to wsdl - Interactions with other groups - GSA-RG participants and also Jim Pruyne (GRAAP-WG) have been responsive - How to engage more people - One issue was the teleconference time slot. It is not so convenient to European participants but it is the only one that has (more or less) worked for US/Japan/UK for some time. - Also need to think how EMS work should relate to DMTF work (cim, profiles, etc) - EMS participants should read up on the material that OGSA-WG got over the summer from CIM & WSDM and then ask engage those groups in a discussion. - Next call: 2 weeks from now Monday