Galaxy workflow management system customization - avoiding duplication of files

[tooony13]

Hi all,

My team is trying to set up an instance of galaxy workflow management system (http://galaxy.psu.edu/) that will launch jobs on our local cluster. We are involved in projects dealing with high-throughput sequencing. We then have to manage a LOT of large files (several Gb each).

When uploading files into galaxy, these files are automatically copied to a folder named "database/files" and they are also sequentially renamed (dataset_1, dataset_2, dataset_3 ...etc..). This name convention is independent of the fact that a file is an input/intermediary/output file.

Copying and renaming files this way is too much time consuming and makes us lose our file structure.

Is there a way to avoid this behavior and that galaxy just remember the file path instead of possessing his own copy ?

If someone here has experience with this tool, any help or useful link would be appreciated.

Cheers,

tony

[dawe]

We solved this using a filesystem with deduplication capabilities (ZFS). To do this you have to switch to an operating system that support it (IllumOS, Nexenta, Solaris).

[jgoecks]

Hi Tony,

I recommend you post your question to (and subscribe to) the galaxy-dev mailing list: http://lists.bx.psu.edu/listinfo/galaxy-dev The Galaxy team reads and responds to email on this list, and it serves a forum for discussion about the technical aspects of Galaxy as well.

FYI, what you want to do is achievable by modifying Galaxy's universe file, but I'm not exactly sure how as it's not my area of expertise.

Thanks,
J.

Emory University
Galaxy Team

Hi all,

My team is trying to set up an instance of galaxy workflow management system (http://galaxy.psu.edu/) that will launch jobs on our local cluster. We are involved in projects dealing with high-throughput sequencing. We then have to manage a LOT of large files (several Gb each).

When uploading files into galaxy, these files are automatically copied to a folder named "database/files" and they are also sequentially renamed (dataset_1, dataset_2, dataset_3 ...etc..). This name convention is independent of the fact that a file is an input/intermediary/output file.

Copying and renaming files this way is too much time consuming and makes us lose our file structure.

Is there a way to avoid this behavior and that galaxy just remember the file path instead of possessing his own copy ?

If someone here has experience with this tool, any help or useful link would be appreciated.

Cheers,

tony