OGSA EMS Teleconference - 25 October 2004 ========================================= * Participants Larry Flon (ISI/Globus) Ian Foster (ANL) Andrew Grimshaw (UVa) Hiro Kishimoto (Fujitsu) Mark Morgan (UVa) Jim Pruyne (HP) Andreas Savva (Fujitsu) Apologies: Chris Smith Minutes: Andreas Savva * Summary of main actions Action: Mark & Andrew to review the document and with a view to finding an appropriate term that could be used to replace 'job' as it is defined right now. And revisit. Action: Andrew and Mark to add 'task' to the list of terminology that needs to be checked. Action: Ravi to provide text for the glossary for job, task and other terms that may be affected by this discussion. Action: Andrew to prepare revision based on comments; Send to Andreas when it is done. * OGSA v1 EMS section public comment review [Numbers refer to an old version of the issues list. See attachement in http://www-unix.gridforum.org/mail_archive/ogsa-wg/2004/10/msg00047.html ] - Andrew and Ravi had already worked through a number of the issues in a previous call. Andrew is preparing a revised version. - Started comment review from the first EMS comment to review revisions. - 50: "usage of service in section title is incorrect": Disagree; the service name describes the function it provides. - 51: "BLAST as only example": Agree; revised to make obvious it is an example of a class of apps. - 52: "non-legacy batch job": Agree that this is not a good term; Revised text. - 53: "Too many unanswered questions": Yes but further work in V2. (It is ok to have unanswered questions in v1 as it is an informative document and describes the issues.) - 54 & 66: Definition of "Job" - It needs to be made clearer that 'job' is modeled as a resource (ws-resource). Text needs some revision. - (That a 'job' may provide (or appear) as a resource to other entities is a separate topic.) - Need to work out differences between job, job manager, job document etc. These are not clear and were a hot topic at GGF12. - A lot of discussion on whether the current definition of job is confusing or misleading. - Agreement that the term 'job' is overloaded and ideally a different term should be used. - Not sure whether this should be v1 or v2 work. - 'job' is not a 'workflow': The current definition is there to make the point that the 'job' is a terminal point; it cannot be decomposed further. - However defining 'job' in this way goes against common usage. Intuitively anything submitted for execution (be it a workflow or 'atomic' job) is a job. Also why not allow for virtualization of this concept. - Another comment [66] highlights this difference in usage between the Requirements and EMS sections. The Requirements section talks about different types of jobs, simple or complex; seems to be common usage so redefining 'job'creates problems. - Note this discussion is just about the term used; it is not trying to restrict or redefine the functionality that the EMS section is describing. - There was a suggestion to change the 'Requirements' to talk about 'entities' rather than 'jobs' but that sounds unnatural (and 'entities' is somewhat of a last choice when no other appropriate term exists). Action: Mark & Andrew to review the document and with a view to finding an appropriate term that could be used to replace 'job' as it is defined right now. And revisit. - 55: "JM responsibility" and "incompatible software": It is an example meant to highlight the JM's role in this specific use case. It does not (and does not need to) cover all aspects. - 56: "State model": Too soon to go to this level of detail. V2 work. - Also note that there is no sub-unit of 'job' (in the sense that subunits are not manageable on their own) - 57: Task and job relation: - Historical background: The 'task' term was not used in the document before submission to public comment. (Ravi's hierarchy that showed task as a sub-units of Job is not part of the document.) There was a problem with the usage of the term 'service' and it was agreed to replace 'service' with either 'resource' (ws-resource) or 'task'. - Agreed that this has caused confusion. And will change or remove usage of 'task' in favor of another term. Action: Andrew and Mark to add 'task' to the list of terminology that needs to be checked. - Also need changes to the glossary to make it consistent with the proposed changes to OGSA v1. - For example, Ravi's definition is that executable aspect of a job is task. Action: Ravi to provide text for the glossary for job, task and other terms that may be affected by this discussion. - 58: "Security interactions": Agree that this is a valid comment but it is v2 work - 59: "translation, validation": a valid comment; but further refinement is v2 work. - 60: Unclear text: Agreed to minor re-write to remove 'plug-points' (replace with 'interface') and clarify. - "Figure not in sync with text": Agreed that it is a valid comments. Decided to color the services that are covered in v1 and add text on what is covered; indicate that other services will be covered in later versions (v2). - 61: Agreed that it is a valid comment: Rewrite text. - 62: Change 'them' to jobs - 63: Revised. - 64: "References...": problem is that some services do not have corresponding sections: - Add references where sections exist; say it is v2 if they do not. - Finish sentence at 'success of OGSA.' - "Logging etc": rephrase as an example and point out relation better. - 65: - Added quotes - Restate JSDL mention as example - Fig 5. is a collaboration diagram: change/revise caption. No need (cannot) add more information at the moment. - 66: See the discussion under 54. Wait for Andrew/Mark's recommendation. Action: Andrew to prepare revision based on comments; Send to Andreas when it is done. * Comments from Olegario for v2 - The comments were submitted as v2 comments. He had already submitted comments on v1. - Agreed to address these in v2 and not v1. * Next call: 2 weeks. Hiro/Ravi to send annoucement on time/agenda.