[kepler-dev] A Workflow related paper and MoML/BPEL4WS
ludaesch at sdsc.edu
Mon Dec 20 13:06:40 PST 2004
Thanks for the note. I think it's worth keeping an eye on these
things. But last time I checked BPEL4WS didn't offer much at all in
terms of scientific workflows.
That's why all scientific workflow systems I know (academic and
commercial) ignore BPEL4WS -- at least for now.
That's not to say that one day several communities might get together
and come up with a joint language for scientific workflows. There are
some "link-up" activities already under way.
When comparing languages it is good to try and come up with some
evaluation criteria. For example, what exactly is the problem being
solved by BPEL4WS?
E.g., MoML solves the problem of capturing Ptolemy/Kepler
workflows. But I view MoML mostly as a "Ptolemy internal" language
since without a Ptolemy environment, a workflow stored in MoML doesn't
tell everything known about it (e.g. to avoid redundancy, certain
information is kept with the Java classes etc).
It would be quite valuable to come up with a "Kepler external"
language that allows to share workflow specifications at various
degrees of refinement (from design, to execution, and even the
associated "provenance workflows", i.e., appropriately customized
traces of executions for reporting purposes).
As mentioned, the there are some efforts eg as part of the Link-Up
project that are related to this.
An interesting project might also be to have a "save-as BPEL4WS"
option for Kepler. It would probably quickly reveal the strengths and
weaknesses of BPEL4WS for our purposes.
And why would one do that? Maybe because there is a free, ideally open
source BPEL4WS "engine" out there. Is that so?
Overall my impression of BPEL4WS is that (a) it's quite bloated (b)
doesn't offer much of what we're interested in (semantic extensions,
interaction support, scheduling support, distributed execution, LSID
or Grid-ID support, provenance, ... ). But again, we should
periodically look what's really real there and then re-evaluate.
just my $.02
>>>>> "JZ" == Jianting Zhang <jzhang at lternet.edu> writes:
JZ> The following paper which will be published in Data & Knlowedge Engineering
JZ> next year seems related to our work on scientific workflow.
JZ> Managing structural genomic workflows using Web services
JZ> Data & Knowledge Engineering, Volume 53, Issue 1, April 2005, Pages 45-74
JZ> Maria Cláudia Cavalcanti, Rafael Targino, Fernanda Baião, Shaila C. Rössle,
JZ> Paulo M. Bisch, Paulo F. Pires, Maria Luiza M. Campos and Marta Mattoso
JZ> They use BPEL4WS to specify their workflow. A quick search tells me that a
JZ> plugin for Eclispe (based on the labelled transition system analyser (LTSA),
JZ> http://www.doc.ic.ac.uk/~hf1/phd/eclipse_project.htm ) is being developed.
JZ> Anyone in Kepler community has the expereinces with BPEL4WS (and its
JZ> supporting graphic tools) can compare them with MoML/Vergi under scientific
JZ> workflow context?
JZ> kepler-dev mailing list
JZ> kepler-dev at ecoinformatics.org
More information about the Kepler-dev