[Pgi-wg] Keep HPC-BP Demo alive with monthly telcons

[Steve Crouch]

Hi Morris, David,

Yes, of course.  In addition (or inclusive of this), it could also
eventually become a vehicle for verification of PGI-WG outputs
(profiles, etc.) as the middlewares adopt them.

There are many possibilities, and as this is already a long email I'll
only focus on a few for now...

As Morris very rightly suggests, the demo needs to be easily usable by
others, with a low preparation overhead, also focusing on stability and
maturity.  I'm planning to upload the demo to GIN-CG's OGF Forge this
week - at least a preliminary upload, with short minor commits next week
to clean up the code base.  It already goes some way to being an
approachable system to download, configure and run, but needs a bit of work.

Since there are many ways in which the demo could be extended, I think
we'll need to consider how best to include any extensions as they are
developed, without sacrificing current functionality, usability and
stability.  Currently, the demo is written in Perl, since it's very easy
to deal with command line invocations and is quick to develop and
extend, but currently the code is quite monolithic, but with partial
comments.  Modularisation of the code base could be a logical first
step, and would facilitate distributed development and extensibility. 
This would enable generalisation of the application used ('Minem' at the
moment), and configuration for utilising other (emerging) specifications
and profiles.  And of course, the general approach of the demo could be
applied to other languages, such as Java, if desired.

As David suggests, it would be a good idea to position it as a test
framework of HPC-BP services in a 'compliant-like' way.  At the moment,
it only utilises CreateActivity / GetActivityStatuses operations - it's
service response parsing is quite naive at the moment (after all, it is
only a demo!).  So, as it stands it's very much an application-oriented
demonstrator that provides just an indicator of basic compliance, but
there's no reason why parsing completeness (from a service response
perspective) couldn't be extended, and also use the other methods on the
BESFactory porttype - TerminateActivities, GetActivityDocuments and
GetFactoryAttributesDocument for example.  The latter two could be done
for each endpoint prior to running the actual application part of the
demo, for instance.

One of the major strengths of this activity has been the willingness of
participants to assist and provide endpoints, as they have for previous
interop testing.  This is really the foundation for its success so far. 
Building on this, and looking to the future, it would be great if we
could try to secure (what do others think?):

 - A commitment from middleware endpoint providers to provide and
maintain endpoints would be ideal for both demonstrating and
development.  To what extent is this practical?
 - For greater credibility, actually having endpoints provided on proper
DCI's also seems compelling (as we have with GridSAM on the UK NGS).

At least from my experience, security trust configuration is always an
issue - ironically perhaps, the basic standards interop testing is quite
straightforward :)  In terms of security, would it be a good idea to
move away from UsernameToken support and move towards X509
certification?  This is more in line with how the infrastructures are
deployed in a production sense, so keeps the security configuration more
simple and more realistic.

My apologies for the long email - but hopefully a starting point for
further discussions.  Any thoughts very welcome!

Cheers
Steve



David Wallom wrote:
> Hi Morris,
>
> I completely agree that this should become a key function of the GIN group,
> I would as I suggested during the standards AH suggest that this could also
> become a sort of compliance system for HPC-BP.
>
> David
>
>
> On 16/03/2010 08:06, "Morris Riedel" <m.riedel at fz-juelich.de> wrote:
>
>   
>> Hi folks,
>>
>>   I had some discussions with several people here at OGF and we thought that
>> it makes sense to have at least one telcon per month under the GIN umbrella
>> to keep the HPC-BP demonstration alive and discuss possible extensions in
>> terms of middlewares/standards/applications. One aim of this would be to
>> make the demo more mature and stable in order to show it at numerous events
>> w/o high preparation efforts.
>>
>> Before I will ask for a specific date I'm open to hear your opinions on this
>> and/or any further suggestions.
>>
>> Take care,
>> Morris 
>>
>> ------------------------------------------------------------
>> Morris Riedel
>> Distributed Systems and Grid Computing Division
>>
>> Jülich Supercomputing Centre (JSC)
>> Info: http://www.fz-juelich.de/jsc/JSCPeople/riedel
>>
>>
>>
>>     
>
>   

-- 
Dr Stephen Crouch, 
Software Architect,
OMII-UK, 
Electronics and Computer Science, 
Room 4067, Level 4, Building 32, 
University of Southampton, 
Highfield, Southampton SO17 1BJ  
Tel: +44 (0)23 8059 8787        EMail: s.crouch at omii.ac.uk 
Fax: +44 (0)23 8059 3045        WWW: http://www.ecs.soton.ac.uk/~stc