BES BoF Minutes -- GGF 13, Seoul Korea Minutes by Mark Morgan In Scope v. Out of Scope ------------------------ * Out of scope - Whether or not the activities themselves can be talked to directly. * What kinds of use cases are in and out of scope? * In scope: Someone says, "I want to run BLAST, right now, on that host, and here is the information about how to do it (what inputs, what binaries, etc.) * Borderline In v. Out of Scope: Someone says, "I want to start a new web service instance on that host." * Out of scope: Someone wants to run a work flow of tasks, or wants to run a blast job, but doesn't know where to run it. * What about being able to get information about what a specific host or container provides? Will likely be in scope, but at the moment is sort of out of scope. In otherwords, we are not doing schedule, but it's the type of thing that may be necessary anyways. * Why assume data staging? Right now, we aren't in fact assuming data staging. Rather, the data group may want to do something different. * What about asynchronous notification? That's in scope Presentations by Other Groups * NAREGI Presentation * JSDL Presentation - Scope of File movement. JSDL says file needs to stage in or out, but not how that happens. It may tell you which services you MIGHT want to use. - Is there any notion that if I am staging something in, can the underlying operating environment say, I already have it? -- Nothing permits or dissallows it. -- If you want to break the staging into pieces, that's fine too. - How does this fit into BES. -- If BES gets a JSDL document that has stage in and stage out and an application to run, then BES may not care. One of the things that the Naregi people have done is put the JSDL doc through a number of transformations. Then, afterwords, the document may not say anything about staging. -- Do we even need to decide that. Can the BES decide on it's own how to handle that. - What purpose does JSDL provide in BES since the location of the place to run has already been decided. -- Yes, but part of what it says is things like: -- it needs 32 G of scratch space, etc. -- Stuff that you will still need to satisfy. -- A BES could be a meta scheduler as well. * GridSAM - What is the difference beween this and DRMAA and SAGA. -- In the interface in GridSAM using JSDL as the language instead of using the DRMA more fine grained API. * DRMAA * Legion GRAM/BES are Bottom Layer Schedule is going to be more agressive then many others. More like OGSA-WG schedule. Done so far - First pass on scope - Looked at several projects -- Seems like a great deal of similarity - Copy of draft charter sent to Stacey. One objective in short period of time is to hammer out draft charter to send to the steering comitte * Scope issues are important One of the things that we brougth up is things that are in scope and out of scope. For example, should we assume anything about the interfaces of job services that get started? We should discuss whether or not the milestones are realistic. In terms of management issues we believe that the working group will have to have telecons at least every other week and possiblly F2F in between GGFs. One of the things that the GFSG has asked is whether or not there is committment for participation which there seems to be. There are pre-existing documents, but they aren't even information documents. Exit Stragety -- when the recommendation document is out, we are done and will dissolve ourselves. Show of hands -- how many are interested in working in this work group? about 20 or 25. We have to get Stacey to turn on the email list for us. Send a note to office and she'll turn on the email list. The name of the email list will be: OGSA-BES-bof@ggf.org Some of the Stakeholders may not want to participate. People who believe that they are likely to use this output of this group. Obvious Stake holders Globus Alliance Naregi OMII Unicore UVa PBS? -- no. Willing to watch, but not participate Platform -- probably a stakeholder. IBM HP DRMAA JSDL National Grid Service, Reality Grids, etc. Business Grid Computing Group Grid Portal Developers Focus on Scientific only? No Discuss who might be potential chairs for this group. Steve Newhouse and Grimshaw have volunteered. Nomination Daren Pulsiphier. First task is to decide on scope. Send out draft charter.