Taverna workflows typically work at a higher level than submitting grid execution jobs with command lines and data staging.
There's some confusion as there are other workflow systems that deal with coordinating grid jobs. Taverna workflows don't specify jobs that are to be done and executing a Taverna workflow does not typically do much local processing. Taverna works in a world of web services and other remote resources - so when it comes down to 'executing a workflow on the grid' there are really two things one are talking about. The first is to have the Taverna workflow engine running as a grid job. Unless we are talking about a workflow that takes quite a long time to run (or you are going to run many instances of this workflow) this does not often give you much, as the 'real' services would still be running somewhere else - all you would have is a grid job which is coordinating web service execution - which would work quite well straight from your desktop. There are several solutions to running a workflow remotely without needing a grid, Taverna Server is one such product we are working on (building on Taverna Remote Execution Service for Taverna 1), which is a RESTful web service that can take a workflow and its input, and perform the workflow execution. You can build your own similar execution by using the Taverna Platform APIs - or even wrap the Taverna Platform APIs as a command line tool that you can then submit as a traditional grid job. (Taverna 1 has such a command line tool, we are in the process of developing a similar tool for Taverna 2) The other approach has to do with the services them selves (the components of the workflow) executing on a grid. This is often much more useful, and it will come down to how you want to expose your services (ie. programs) to Taverna. The roughest and most general cut to this would be to have a web service that can take any command line and submit it as a grid job. Of course the software would have to be pre-installed on the cluster, and the user would have to know both the paths, command line syntax and parameters. GridSAM is one such product that gives a web service for submitting jobs. There have been developed closer integrations along the similar line, like gLite support by Ketan Maheshwari (which you will find on this list) - but which can also deal with file transfers up and down for you. Also see: http://www.mygrid.org.uk/mygrid-in-use/mygrid-use-on-a-grid/ http://www.mygrid.org.uk/tools/taverna/taverna-1/taverna-documentation/taverna-frequently-asked-questions/taverna-and-grids/ Then you can go one step further and developer a real web service specialized for your program. Tools like Soaplab and gRAVI can help you do this. The advantage of this is that you can describe the 'real' parameters of the service (if you want, even with typed inputs and outputs, not just strings), and the workflow designer no longer need to deal with command line parameters and file staging. You can at this stage also probably get rid of the authentication requirements as you have limited the service to only execute your selected program. In addition your service can be publicly shared and described on websites like BioCatalogue - and workflows using your service can be reused by others by sharing it on MyExperiment. This is the real motivation for the service oriented approach that has been driving software like Taverna, and opening up for unexpected new approaches as different services come together. On Thu, Jul 30, 2009 at 18:49, Jordan Mendler<[email protected]> wrote: > Is it true that Taverna does not have direct functionality to submit jobs to > clusters and monitor the cluster jobs? Based on what I can find, it seems > clustering is done by writing your own cluster layer and then executing a > command a remote host with a script to submit the jobs. Is this correct? > > Relevant threads: > http://markmail.org/message/eeto5ygyo4zai4s4?q=list:taverna+-type:checkins+sun+grid+engine > http://markmail.org/message/xyskxsln6u5ragaz?q=list:taverna+-type:checkins+cluster > > Thanks, > Jordan > > ------------------------------------------------------------------------------ > Let Crystal Reports handle the reporting - Free Crystal Reports 2008 30-Day > trial. Simplify your report design, integration and deployment - and focus > on > what you do best, core application coding. Discover what's new with > Crystal Reports now. http://p.sf.net/sfu/bobj-july > _______________________________________________ > taverna-users mailing list > [email protected] > [email protected] > Web site: http://www.taverna.org.uk > Mailing lists: http://www.taverna.org.uk/taverna-mailing-lists/ > > -- Stian Soiland-Reyes, myGrid team School of Computer Science The University of Manchester ------------------------------------------------------------------------------ Let Crystal Reports handle the reporting - Free Crystal Reports 2008 30-Day trial. Simplify your report design, integration and deployment - and focus on what you do best, core application coding. Discover what's new with Crystal Reports now. http://p.sf.net/sfu/bobj-july _______________________________________________ taverna-users mailing list [email protected] [email protected] Web site: http://www.taverna.org.uk Mailing lists: http://www.taverna.org.uk/taverna-mailing-lists/
