RE: Workflow

["Tomasz Haupt" <haupt@erc.msstate.edu>]

> Yes, I think we do want to call this workflow.

Personally I can live with it. What bothers me it that other people had
problem with it. So let us hear from them. If nobody responses, we will use
"workflow".

>> 2. Is the "sequencing" not too restrictive?

> True, GALE is strictly serial dependencies now, and I think the work that
> you and others have done would be very useful.

OK, let us work on it.

>>
>> 3. It seems to me you are suggesting an enumeration of "atomic tasks":
>>   computation, resource query, data transfer, ..., etc. Again, is it
>>   not too restrictive? Actually I have several problems with that.


> When I started designing GALE I struggled a lot with whether to define a
> general task or activity. One of my design goals was to be able to create
a
> "grid script" that non-IT specialists would be able understand. To me,
that
> meant defining specific activities like DataTransfer and Computation. Most
> of our users will never see the XML, but like always there are those that
> hate GUIs and prefer commandline tools and editing text files directly.

I think we are talking about two different things: a task/workflow
descriptor and its user interface (not necessarily GUI). For the end user
there will be always well defined "modules" or "atomic tasks" such as
DataTransfer or Computation of some sort. Those are just names associated
with the functionality of the "module". My problem is how do we define the
functionality of the module? I would suggest that the fuctionality be
defined through a sort of "application descriptors". Keep in mind that a
Computation may be a quite complex in terms of configuration, creating
scripts etc.

The bottom line is that I would prefer to separate definition of how to
process a single node in a computational graph from definition of
relationships between the nodes. I think that the workflow document should
concentrate on the latter. The former, IMHO, belongs to application
descriptor. After all, one may want to run a single application in a stand
alone mode and it would be redundat to have different descriptors depending
whether the application is a part of a workflow or not.

Tomasz




I agree that we need a more sophisticated transition mechanism based on
events.