[SDM-SPA] Re: [kepler-dev] Re: [seek-dev] notes from workflow workshop

[Bertram Ludaescher]

Efrat:

Thanks very much for sending this nice summary to the various groups!

Bertram


>>>>> "EJ" == Efrat Jaeger <efrat at sdsc.edu> writes:
EJ> 
EJ> I was checking two of the workflow environments; Triana (Matthew Shields at
EJ> Crdiff University) and Taverna/MyGrid (European Bioinformatics Institute, IT
EJ> Innovation and the Human Genome Mapping Project Resource Center) and below
EJ> are my notes.
EJ> 
EJ> TRIANA is very similar to Ptolemy, both in the GUI part and in the src, but
EJ> it has a single computation semantic (reacting to signals).
EJ> 
EJ> It has a nice web services environment (although most of the their features
EJ> are already being added to Ptolemy/Kepler by Ilkay); a user can load any web
EJ> service from a specific repository (limited to a single repository) and the
EJ> interface creates a folder for each service added and adds its methods as
EJ> actors (the stub source of a web service is automatically generated by
EJ> WSDL2Java and is then compiled and loaded). The user can also retrieve a web
EJ> service outside the repository by providing a specific WSDL URL.
EJ> 
EJ> In order to discover web services input/output types, Triana has two WStypes
EJ> actors which automatically detect the type of its connected web service
EJ> method actor as it is being linked together (including sequence types) and
EJ> thus the user can add the input/output ports accordingly. With those
EJ> automatic type generators the input can be entered either as parameters
EJ> (manually) or as input port from a previous task (the output as well can be
EJ> either displayed or forwarded).
EJ> 
EJ> The interface also allows users to add new units/actors (similar to Chad's
EJ> new actor feature). It is possible to edit an actor's source code from
EJ> within the interface and then compile and run the new actor from within the
EJ> GUI (although I'm not sure of the stability of this feature and of how
EJ> recommended it is to allow the workflows users to edit the source code,
EJ> which can be saved automatically).
EJ> 
EJ> 
EJ> In TAVERNA everything is based on web services.
EJ> 
EJ> The taverna interface consists of several different view windows. The actors
EJ> are web services and are listed under the available services panel. Users
EJ> can load services and nested workflows from the web.
EJ> 
EJ> A model is described as an XScufl (XML Simple Conceptual Unified Flow
EJ> Language), which is similar to MOMLs. As in Ptolmey the interface plays a
EJ> role as a model (XScufl) editor, when a user adds an actor to the workflow
EJ> diagram it automatically generates the XScufl. A workflow is displayed on a
EJ> static graphical view and also on a tree view of the model which lists all
EJ> the components and allows the user to specify the connections between them
EJ> and define input and output characters (unlike Ptolemy the connection are
EJ> not linked in the graphical view, but in the tree view, the graphical view
EJ> is then automatically refreshed to show the links) .
EJ> 
EJ> In order to run the workflow there is a workflow input panel in which the
EJ> user can enter the input data, either a text or a file, and then the output
EJ> is displayed on a result window. The possible output types are the same as
EJ> provided by web services. It is viewed on the screen and can also be saved
EJ> to a file.
EJ> 
EJ> This environment consists a single computational model. The execution is
EJ> integrated via an encantor which specifies coordination constraints, but it
EJ> is very limited for now and is hardly used.
EJ> 
EJ> 
>> 
>> >>>>> "CB" == Chad Berkley <berkley at nceas.ucsb.edu> writes:
CB> 
CB> Hi,
CB> Attached are a few notes that I took at the Workflow workshop in
CB> Edinburgh.  These are the projects that I thought were interesting and
CB> had something in common with one or more of our projects.  Some of
EJ> them
CB> you have, no doubt, heard about already.  Feel free to email me with
CB> any questions...maybe I'll be able to answer them.  All of the given
CB> presentations with contact info can be seen here:
CB> http://www.nesc.ac.uk/action/esi/contribution.cfm?Title=303
CB> 
CB> chad
CB> 
CB> narada brokering: naradabrokering.org
CB> stream oriented workflow: xgsp web service mcu arch.
CB> 
CB> 
CB> chimera (argon):
CB> virtual data language - describes data transformations (lineage
EJ> tracking)
CB> transformation - template of program invocation
CB> derivation - stores past and future (actual function call)
CB> invocation - how derivation was executed (log)
CB> 
CB> MyGrid
CB> "Taverna" - build, edit and browse workflows
CB> import services
CB> integrated execution via Enactor
CB> Taverna Workflow Workbench - nice gui workflow editor
CB> supports control flow and data flow
CB> built in provenance control
CB> written in java
CB> taverna.sourceforge.net
CB> freefluo.sourceforge.net
CB> 
CB> Geodise (engineering)
CB> allows the use of scripting within workflows
CB> map scripts onto workflow components
CB> workflow tool includes semantics and integrated ontologies
CB> using protege
CB> intelligent workflow monitoring and advice (rule based to case based
EJ> in real time)
CB> 
CB> Choreology - OASIS WS BPEL TC
CB> BPEL - XML language for defining behaviour of a process
CB> provides web services
CB> uses web services
CB> everything is a web service (no external interactions)
CB> 
CB> WS-CDL - Web Services Choreography Description Language (w3c)
CB> goal: declarative language that describes collaborations of web
EJ> services
CB> participants by defining their externally observable behavior as
EJ> part of an
CB> interorganizational transaction
CB> choreography - the observable public exchange of messages
CB> orchestration - private executable business process
CB> BPEL can be used to make choreography and orchestration work
EJ> together
CB> web servies participants interact and align their shared information
CB> set of interactions are combined to form a behavioral unit
EJ> (choreography)
CB> 
>> _______________________________________________
>> kepler-dev mailing list
>> kepler-dev at ecoinformatics.org
>> http://www.ecoinformatics.org/mailman/listinfo/kepler-dev
>>