[sem-grd] Re: GGF14 Semantic Grids and the Agents Community

[Kashif Iqbal]

Hi all,

First of all thanks to Dave for providing food for thought on some
interesting issues.

I would like to trigger some discussion on the following point on the
integration issues of various technologies (Grid services, Web Services,
Agents, Semantic Web Services..)

>- People in the research community are looking at agents and web 
>services and at semantic web services.  What happens when working
>with grid services instead? e.g. Are the existing solutions affected 
>by the state-handling?  Do grid services go some way to providing 
>agent functionality beyond web services?  Are OWL-S/WSMO/WSDL-S
>neutral about the extra functionality in Grid services?

1. In one of the earlier discussions on the integration issues of agents and
grid computing, I have pointed out that most of the solutions that aim to
integrate agents and web services, they usually end up in bridging the gap
between the communication protocols. OR atleast this is what I have
experienced. 

Probably we need some effort like JAS (Java Agent Services) which ended up
prematurely, unfortunately. JAS like functionality will be needed in order
to fill this gap and provide interoperability with existing service oriented
architectures like (web services and Grid) 

2. Agent Community has a lot of experience in Communication and Interaction
protocols that provide the basis for autonomy and negotiation in an Agent
Platform. And current research inititaives in web and grid world clearly
depict that both web and grid world also require negotiation and
communication languages. SLAs is one of the step in this direction. Web
service choreography is another for defining interactions with a given web
service. People in semantic web services community are now developing
choreography languages to define complex interaction patterns.

In agent paradigm, an agent cannot directly invoke the method of another,
rather they have to exchange ACL messages which can ultimately leads to
negotiation and interaction before invoking another agent behavior or this
model can be used to acquire a resource or invoking a method of a service,
or forming a VO in Grid. Current Web Services standards lack this model of
interaction and invocation.

But no doubt web services need interaction and negotiation and Choreography
languages and SLAs specs for negotiation are steps forward in this
direction.

Now the question arises that if agents characteristics like communication
and negotiation protocols are adopted by web services without any feedback
from agent community, how we will end up in the integration of these two
technologies?

3. As far as state handling and existing solutions are concerned, existing
developing systems (those which I know, and I am working with) like WSMX
(Web Service Modelling Execution Environment) and Adaptive Services Grid
platform which is still in design phases will be using WS-Addressing for
handling state. GT4 has also implemented WS-addressing. WSRF.Net is another
example.

4. As far as WSMO and Grid service are concerned. We are planning to develop
GSMO (Grid Service modeling ontology) to capture concepts from the Grid
world and to describe Grid Services comprehensively. I have proposed this in
GGF-12 and feedback was to get some use cases from the Grid community and if
possible then a GGF working group should be established in order to allign
with Grid standards. (May be Dave can suggest how should we proceed with
this?)

These are some of my personal thoughts, please correct me if I am wrong
somewhere.

Best regards,
Kashif



-----Original Message-----
From: owner-sem-grd at ggf.org [mailto:owner-sem-grd at ggf.org] On Behalf Of
David De Roure
Sent: Wednesday, June 29, 2005 7:35 PM
To: Semantic Grid Research Group
Subject: [sem-grd] Re: GGF14 Semantic Grids and the Agents Community

Hello

Some thoughts from the discussion at the end of this morning's
session:

- The world we describe, where the decision about exactly which
service is to be used is taken out of the hands of the users, may 
not be what all users want.

- It was apparent that the agents approach is more "loosely coupled" 
than the grid services approach appears to be at this time, and may
therefore be particularly beneficial at the inter-grid level.

- How to engage the agents and grid communities?  In agents there have
been very successfuly competitions / challenges (Jon Dale ran an Agent
Cities one).  Could we do a challenge on the Grid?  This would require
some Grid services.  I will take this issue back to GGF - i.e. will 
GGF provide/coordinate/list Grid services for community use?  It also 
requies a scenario(s) (which must be motivating!)  Suggestions?

- People in the research community are looking at agents and web 
services and at semantic web services.  What happens when working
with grid services instead? e.g. Are the existing solutions affected 
by the state-handling?  Do grid services go some way to providing 
agent functionality beyond web services?  Are OWL-S/WSMO/WSDL-S
neutral about the extra functionality in Grid services?

Please add to this anything I've forgotten (I'm sure there's something...)

Incidentally, I heard from Julie on the GGF operations side that we 
had over half the registered participants at GGF in the community track
this morning - this is a clear success for the community track approach!
I welcome comments on this that I can feed back into the GGF community
council.

Thanks

-- Dave