[MOBY-dev] Moby Collections in Taverna

[Edward Kawas]

Hi Dirk,

The output port behaviour has been cloned for collections with a new port,
Collection 'As Simples'. So that port now takes the n items in a collection and
produces n simples that Taverna can iterate over.

However, if that port isn't used and a collection is passed to a service that
expects a simple, I transparently break apart the collection with n items into n
invocations containing a simple and then invoke the service n times, rather than
one time with a n invocation message. This was done, because I noticed that
there were time out issues with very large collections.

I haven't committed my changes in the Taverna cvs yet, so the only way to try
these changes is to download an updated taverna-1.3.1.jar file from
http://bioinfo.icapture.ubc.ca/ekawas/jars/taverna-1.3.1.jar. I also have some
example workflows at http://bioinfo.icapture.ubc.ca/ekawas/workflows/. Once I am
ready, I will commit my changes and hopefully in the next release of Taverna
(very soon I hear) the plugin will be updated.

Thanks,

Eddie

> -----Original Message-----
> From: moby-dev-bounces at lists.open-bio.org 
> [mailto:moby-dev-bounces at lists.open-bio.org] On Behalf Of Dirk Haase
> Sent: Thursday, April 13, 2006 12:58 AM
> To: Moby Developers
> Subject: [MOBY-dev] Moby Collections in Taverna
> 
> Hi all,
> 
> I noticed that the handling of Moby collection outputs has 
> changed in Taverna. 
> Formerly (until version 1.2), a collection of n objects was 
> processed with n invocations of the following step in the workflow.
> 
> Now (version 1.3.1) this is done with only one invocation 
> with n distinct queries (mobyData blocks). That's nice, 
> because it reduces overhead. The downside is that for large 
> collections this will inescapably lead to timeout problems. 
> For our blast services for example I observed a limit of 
> about 20 queries per invocation.
> 
> I also found that the old multi-invocation behavior is still 
> accessible by using the non-moby 'output' port. However I'm 
> not sure if this is really the intended way...
> 
> I think the optimal solution would be to have a 
> 'max-number-of-queries-per-invocation' parameter to be set in 
> the workflow definition. Is that feasible?
> 
> Regards,
> dirk
> 
> -- 
> 
> ----------------------------------------------------------
> Dirk Haase		            phone +49 89 3187 3583
> http://mips.gsf.de/~haase           email   d.haase at gsf.de
> _______________________________________________
> MOBY-dev mailing list
> MOBY-dev at lists.open-bio.org
> http://lists.open-bio.org/mailman/listinfo/moby-dev