[ogsa-wg] Re: [saga-rg] Re: OGSA EMC BoF on Simple job submission
Hiro Kishimoto
hiro.kishimoto at jp.fujitsu.com
Tue Feb 8 12:06:09 CST 2005
John's email bounced.
----
Hiro Kishimoto
From: John Shalf <jshalf at lbl.gov>
Subject: Re: [saga-rg] Re: OGSA EMC BoF on Simple job submission
Date: Tue, 8 Feb 2005 09:37:14 -0800
To: Bill Nitzberg <bill at computer.org>, ogsa-wg at ggf.org,
Hrabri Rajic <hrabri.rajic at intel.com>,
Simple API for Grid Applications WG <saga-rg at ggf.org>,
Stephen Pickles <stephen.pickles at man.ac.uk>,
Tom Goodale <goodale at cct.lsu.edu>
On Feb 8, 2005, at 8:46 AM, Tom Goodale wrote:
>> Would
>> SAGA-RG people like to engage OGSA EMC and possibly co-sponsor a
>> Simple
>> job submission WG? This could be a nice opportunity to tie in the
>> classic API and the WSDL approach.
>
> I'm not sure how relevent this would be, unless you are thinking of a
> direct mapping of the SAGA API into WSDL, which may not make sense,
> and would probably constrain the new group too much, however we should
> certainly make sure we have some people at the BoF.
Well, this should be cause for further discussion. The application
programmer point of view on this will be "I don't really care what's
underneath as long as it does what I ask of it when I call the right
subroutine". So the WSDL mapping is more on the implementation side of
things. That being said, having a working implementation is quite
important. So it may be very interesting to pursue as a path to
implementation (albeit not the only path). I think the concern on the
SAGA side is that the WSDL interface will be adjusted to accommodate
the semantics defined by the application use cases rather than the
other way around. But this sounds pretty reasonable to me.
>> One thing to consider: Do we really need APIs in SRM, APME, and now
>> in
>> ARCH areas?
>
> I'm not sure I understand this question ? We have APIs for the
> functional areas we identified from our use cases.
Just to echo this part of Tom's note, the API's are driven by the use
cases. The use cases thus far didn't describe any good SRM
applications (or at least SRM-like requirements were not described in
enough detail for us to act upon). It could just be that the apps
people are so bogged down in remote file access issues that they don't
realize they need good Storage Resource Management systems yet. If we
get some well described SRM, APME, or ARCH use cases, that will
motivate development of APIs in those areas. (its not that those areas
are not important)
-john
More information about the ogsa-wg
mailing list