[SAGA-RG] OGF 18 - SAGA Minutes

Dieter Kranzlmueller dk at gup.jku.at
Tue Sep 19 11:50:13 CDT 2006


Hello All,

thanks for the minutes and sorry for my absence this time due to another 
meeting. I am glad to hear about the discussions you had and some points 
sound interesting to be followed-up in the near future.

I also noticed one particular point about streaming and tunneling, involving 
SSH. If you are interested, take a look at http://www.gup.jku.at/glogin, 
which is a little tool that helps with the implementation details of such a 
solution.

Regards,
Dieter


----- Original Message ----- 
From: "Pascal Kleijer" <pascal.kleijer at knd.biglobe.ne.jp>
To: <saga-rg at ogf.org>
Sent: Tuesday, September 19, 2006 3:34 PM
Subject: [SAGA-RG] OGF 18 - SAGA Minutes


> Hello all,
>
> Here are the minutes of the two sessions held at OGF in Washington DC.
>
> **********************************************************************
>
> Session: SAGA I
> Time: Tuesday 12 September 2006, 10:00-11:30 AM
> Place: Washington Convention Center - Room 155
> Attendees: ~15
> Steering: Shantenu Jha
> Notes: Pascal Kleijer
> Phone: Andre Merzky
>
> Administration:
> ---------------
> IPR
> SAGA Status and history for newbie.
> State of the RG and the WG, API v1.0 submission after GGF 18.
>
> SAGA Look & Feel:
> -----------------
> Andre: Simple and quick overview of the Look and Feel (L&F) aspects of the 
> specification. What is in and what not.
>
> Ron: Does Java implementation interacts with GT4?
> Andre: CoG might be interacting but GT4 is not indented.
> Shantenu: Might be in the future by some group in the UK.
>
> Andre: Brief explanation of the slides. A more detailed explanation of the 
> Interface & Class hierarchy. Discussion first focuses on the L&F, and then 
> the functional API.
>
> Q: Where is the document in Grid forge?
> A: The SAGA-Core-WG documents section should contain it; it is found in 
> the SAGA-RG documents section.
> TODO: Synchronize both projects in Grid forge to cross point the relevant 
> documents.
>
> Q: What happens if a method in the specification doesn't have an 
> underlying implementation?
> A: Then the implementation is not compliant with the spec.
> Q: What then?
> A: We can throw an exception. This is in the spec document, the General 
> Discussion section.
>
> Q: Discussion of the BES state model and SAGA is there compatibility? A: 
> yes, there are some additional states.
>
> Discussion of COMPLIANT but INCOMPLETE, NOT COMPLIANT. Can be found in the 
> document.
>
> Q: L&F is not an appropriate name; something better (bit mis-leading)?
> R: "Foundation" or "Principle" might be a better? Needs to think about it! 
> One proposal is to replace it with 'Principles of Foundations'
> TODO: Fire this issue on the ML for future discussion.
>
> Q: Can we break down into namespace or blocks to support sub-sets to be 
> implemented? This could allow partial implementation.
> R: Yes this is already possible even if there is no clear namespace 
> definition.
>
> TODO: Andrew Grimshaw pointed: SAGA should send experiences to OGF SC 
> about compliance.
>
> TODO: FIXME must be removed before the document is sent to the OGF SC.
>
> Q: Is this document supposed to help the developer or make then have head 
> aches?
> R: No we try to help them as much as we can.
>
> TODO: Get feedback is important to improve it, seek feedback as much as 
> possible.
>
> Q: What middleware is supposed to be supported?
> R: C++ version should support GT4 in the future.
> e.g., OMII implementation supports their stack, though that might not be 
> so in the future.
> Remark: Vendors support needed to develop bindings to their platform.
> Andrew Grimshaw: Go to customers, all vendors were involved.
>
> This is a problem that OGF in general has to help tackle. It is not only a 
> SAGA task. OGF SC needs to get involved as well.
>
> Why not use the MPI model? Take the vendors together, it that feasible 
> right now?
>
> Customer: Stable API is a must, possibility to switch between 
> implementations.
>
> SAS representative: Will be useful if it provides all functionality
> (i.e., don't want to have to depend on different APIs). Equally don't want 
> statically linked libraries!
>
> Go directly to the customers (e.g., Banks, Boeing, TeraGrid...), not just 
> vendor, to ask them if they need it and how they need it. Use the 
> customers to push ISV to support SAGA.
>
> TODO: Shantenu Jha needs to take contact with several people for future 
> discussions. Business card taken: SAS, Intel, Qualcomm & GSI
>
> Chris: better framework to help quick implementation for sub-sets of the 
> API. Platform could focus on the just Job Mgmt
>
> Adoption:
> ---------
> Steven: SAGA should be pushed in the OGF technical strategy committee 
> (document in the pipeline). See the TSC.
> - SAGA finds important mention in the OGFs Tech Strategy committe
>
> Q: SAGA in GIN?
> A: Steven Newhouse mentions that there is considerable debate ongoing
>
> Marketing material to promote SAGA is a necessity to make it pushed 
> forward. It should not be an API technical document for developers but a 
> manager type document.
>
> Q: Who should SAGA approach first: Vendor, Customer or ISV?
>
> EGEE Earth science at the Frauenhofer Institute (SCAI) is interested in 
> testing SAGA.
>
> TODO: Seek contact with them to see what can be done.
>
> Andre: GT4, EGEE (gLight), NAREGI, etc. (big projects) should provide a 
> binding to help push the project.
>
>
> Steven: SAGA is client side stuff thus we should promote that in the 
> marketing.
>
> (SAS) provide a new use case about their needs. Grid Monitoring is 
> important to them.
>
> TODO: Send use case canvas to them (Shantenu Jha)
>
> Next session expectations:
> * Streaming
> * Messaging
> * Next step
> * Marketing material
>
> There were offline discussions with SAS representative and with RAL 
> representative about information services.
>
> Session closed at 11h30.
>
> **********************************************************************
>
> **********************************************************************
>
> Session: SAGA II
> Time: Tuesday 12 September 2006, 13:30-15:00 PM
> Place: Washington Convention Center - Room 155
> Attendees: ~12
> Steering: Shantenu Jha
> Notes: Pascal Kleijer
> Phone: Andre Merzky
>
> Administration:
> ---------------
> Agenda Bashing
> IPR
> Session rescheduling
>
> Streaming:
> ----------
> Remark: Typo in section 3.12 GWD-R.72: "627" at the intro of the 
> paragraph.
>
> Q: Juelich VISIT group, in their implementation wants to explore 
> extensions in the stream API.
> R: Some can be covered in the messaging API.
>
> Q: Is their any group using the API?
> R: Yes at LSU is working on visualization apps using the streams.
>
> Their use case did not cover their new needs. SSH tunneling might be 
> necessary inn the future.
>
> TODO: Juelich can be put in contact directly with John Shalf for more 
> details and getting things moving on.
> Update: Done (Shantenu Jha)
>
> Messaging:
> ----------
> Intro from Andre about this API.
>
> Q: Did the API solve the visualization uses cases?
> R: Yes it did. Large messages support causes some problems at the 
> implementation level but manageable.
>
> Q: How can we do bi-directional streaming through one tunnel?
> R: This is an implementation issue. The API doesn't care.
>
> Q: Any thoughts on this?
> TODO: Put things on the ML for more detailed feedback. Clean up stuff then 
> post.
>
> Next Step:
> ----------
> Platform: Symphony presentation at GGF 17. Platform is looking for an API 
> for their product (Symphony).
>
> Q: Is any other person interesting in this extension? Possible solution is 
> working on the SAGA L&F from Platform side.
>
> Q: Performance important?
> R: Important to critical.
> TODO: Remind Chris to post his proposal on the ML before the next call.
>
> Marketing:
> ----------
> TODO: get contact with Steven Newhouse and Andrew Grimshaw about what to 
> put in the next marketing material flier. Pascal takes the job.
>
> CPR:
> ----
> Quick overview of what it is!
> TODO: Send CPR document to the SAGA ML (Derek's job).
> Update: Done
> Closing Remarks: Any contribution for getting the SAGA out to the outside 
> is welcome.
>
> Session close at 14:30.
>
> **********************************************************************
>
> k-pasukaru at ap.jp.nec.com
> --
>  saga-rg mailing list
>  saga-rg at ogf.org
>  http://www.ogf.org/mailman/listinfo/saga-rg
> 




More information about the saga-rg mailing list