Scheduling and WWW - some ideas and a request for discussion


Subject: Scheduling and WWW - some ideas and a request for discussion
From: Wolfram Conen (conen@conan.wi-inf.uni-essen.de)
Date: Fri Feb 23 1996 - 19:46:43 MET


Dear reader,

in the course of organising a workshop ("Web infrastructure for collaborative
applications" at WETICE, June 19-21, Stanford, CFP at http://conan.wi-inf.uni-
essen.de/WETICE, please, have a look, or, even better, contribute.), I
collected a few ideas on how WWW might be used in a scheduling context.
If you are interested in this question, please, have a short glance at the
two emials I enclosed below. Any comments/ideas/experience reports etc. are
more than welcome. (If I receive them directly, I will summarize here and/or
make them available via WWW)

Here is a short summary:
Firstly, the following question is considered: Wouldn't it be nice to offer a
possibility to test components of business applications (as scheduling
components) in a "working" application environment? Secondly, the use of
WWW in an environment where distributed scheduling/planning seems to be
adequate, is discussed. Thirdly, the possibility to make a factory simulator
available (via WWW/Internet) to enable people to plug-in their reactive
scheduling and control algorithms is mentioned.

Here the text follows (from an email to C. Le Pape and his answer)
-------------------------------------------------------------------
Just a few words about WWW/Scheduling:
Well, one of my initial motivations to propose such a workshop was the
following idea:
(1)
Wouldn't it be nice to offer a possibility to test components of business
applications (as scheduling components) in a working application environment?

Why not use the WWW as a basis for such an distributed, open, and active
system? (One would need a few more coordination mechanisms to coordinate
active Web pages (written in Java or another language)).
WWW is widely available and would allow a smooth transition from intra-
company to inter-company information transfers etc.
Ideally, one should build a basic engine which supports components in
integrating themselves into the functional/data/organisation/process model
the engine supplies or manages. One would need semantically clear interfaces
between the components and the engine. The engine could hide detailed
information about the underlying models. Components and the engine should be
able to carry out dialogs about the services offered/needed, about the
necessary/optional relations to the models etc. To state it plain:
the engine should do the "integrational work" for the components - it
should translate/transform information flows and it should know which
parts of the models and which components are needed to test/operate a new
component. This could allow researchers to test components (such as
scheduling algorithms) in a more complete business environment and this
would allow easier benchmarking etc.
The most interesting benefit might be that a "good" engine would allow to
develop a more or less complete business application completely distributed -
every interested researcher could contribute components (which have to
obey certain communication rules and be able to interpret the dialog with
the engine, i.e. they have to be able to offer their services semantically
correct, to inform the engines about the necessities of their integration
into data/process/org model etc.) and test his components in a complex
environment...well, enough about that.

Another possibility would be to allow a component designer to develop
data/process/org models and an (enhanced) WWW server would maintain this
models for him such that he could test his ,e.g. scheduling approach, in
a self-designed (small) environment without the need to maintain all the
models himself (this is a rather trivial idea - not to easy to implement on
the other hand (modeling tools etc.) ;)
 
That was the initial idea. Another possibility to connect WWW and scheduling:

(2)
a) If you have a situation of distributed scheduling (supplier/producer,
multiple workshops or whatever) and if you have strong needs to integrate
your scheduling with other functional branches of the firm or other stages
of planning (e.g. MPS), and if you believe in mixed-initiative scheduling
(leitstand ?!), it might be attractive to try to develop a WWW-based
application to integrate the different needs/goals by allowing different
participants of the planning/scheduling process to travel through the
constraints/plans/schedules others have to obey or have generated, to
somehow coordinate the distribution of new informations (new plans, new
constraint states, new constraints etc.), to give users on all "ends" of
the planning process a unified mechanism to enter and post new
constraints/informations to the other participants etc. (focus on distribution
and integration of goals/problem solving and the userr interface necessary to
supprot the integration)
b) This would be especially attractive, if you try to adopt a cash-flow-
oriented model of goal (resp. constraint) integration (for example,
inside a market-model approach to goal satisfaction/coordination), because
this would offer you a rational instrument for prioritizing goals and
constraints (because a modeler or designer of such a planning/scheduling
process is allowed to think/model strictly local -individual preferences and
production functions and a certain market mechanism could guarantee a
pareto-optimal solution...but that is another story)

But both has not too much to do with WWW - WWW would only be a vehicle to
explore the needs resulting from coordination of goals/activities in
a distributed and open system again. Nevertheless, with a good reason:
WWW IS widely used - and everybody IS trying to use it (and Java etc.) - so
THIS open environment could help distributed solutions to emerge as THE
standard solution to complex problems and to receive the attention they
deserve ;)
...

(3) (C. Le Pape)
...
Actually, another interesting possibility would be to have a simulator
of a factory available on the net, with a language enabling people to
plug their own reactive scheduling and control algorithms ... Maybe this
would be also be worthwhile to the community.

That's it.
Best regards,
        Wolfram

-- 
Wolfram Conen                             	conen@wi-inf.uni-essen.de
			 	  http://conan.wi-inf.uni-essen.de/conen/
-------------------------------------------------------------------------
Wirtschaftsinformatik der Produktionsunternehmen |  Tel.: (0201) 81003-58
Universitaet-GHS-Essen, 45117 Essen, Germany     |  FAX:  (0201) 81003-67



This archive was generated by hypermail 2b25 : Fri Mar 03 2000 - 16:17:08 MET