help
help > RE: Update to Standalone CONN?
Apr 18, 2017 06:04 PM | Alfonso Nieto-Castanon - Boston University
RE: Update to Standalone CONN?
Hi Damon&Eric,
Regarding CONN standalone availability I will update this week the compiled unix version to the latest release.
Regarding MCR, standalone versions should work perfectly fine within a cluster environment that has Matlab Compiler Runtime (MCR) installed/available (instead of Matlab). MCR is publicly available and it does not uses/requires Matlab licenses. All parallelization options automatically adapt to the environment of the caller CONN version, so all you have to do is make sure that you run CONN using the same version that you want your cluster nodes to use (e.g. running your analyses using the standalone version will automatically use the same standalone version in all submitted jobs).
To be precise, the following cases are automatically detected/handled by CONN:
1) when running CONN from Matlab; submitted jobs will use calls to your matlab executable (e.g. /bin/matlab); if there are multiple Matlab, CONN, or SPM versions installed in your cluster, individual nodes will use the same versions as the ones that were used to submit the jobs
2) when running standalone CONN (either the one shared here at NITRC or one compiled on your own using Matlab Compiler); submitted jobs will use calls to "conn" or "run_conn.sh" executables (the first one found in your system's path); if there are multiple standalone-CONN or MCR versions in your cluster, individual nodes will use the same versions as the ones that were used to submit the jobs
3) when running standalone SPM (only "full batteries" SPM standalone version, which includes popular toolboxes with CONN among them; this will be released some time soon by the SPM group); submitted jobs will use calls to "spm12" or "run_spm12.sh" executables (the first one found in your system's path); if there are multiple standalone-SPM or MCR versions in your cluster, individual nodes will use the same versions as the ones that were used to submit the jobs
@Damon: Let me know if this answers your questions or if I am misinterpreting (perhaps you were suggesting that it might be helpful to let the user decide whether submitted jobs will use matlab or standalone cluster implementations, or something along these lines?)
Last, regarding changes across versions, the file change.log in each release contains a detailed list of all changes in each version (the ones marked with "*" are the main ones that you want to pay special attention to in case you are updating in the middle of some already-existing CONN project)
Hope this helps
Alfonso
Originally posted by Damon Lamb:
Regarding CONN standalone availability I will update this week the compiled unix version to the latest release.
Regarding MCR, standalone versions should work perfectly fine within a cluster environment that has Matlab Compiler Runtime (MCR) installed/available (instead of Matlab). MCR is publicly available and it does not uses/requires Matlab licenses. All parallelization options automatically adapt to the environment of the caller CONN version, so all you have to do is make sure that you run CONN using the same version that you want your cluster nodes to use (e.g. running your analyses using the standalone version will automatically use the same standalone version in all submitted jobs).
To be precise, the following cases are automatically detected/handled by CONN:
1) when running CONN from Matlab; submitted jobs will use calls to your matlab executable (e.g. /bin/matlab); if there are multiple Matlab, CONN, or SPM versions installed in your cluster, individual nodes will use the same versions as the ones that were used to submit the jobs
2) when running standalone CONN (either the one shared here at NITRC or one compiled on your own using Matlab Compiler); submitted jobs will use calls to "conn" or "run_conn.sh" executables (the first one found in your system's path); if there are multiple standalone-CONN or MCR versions in your cluster, individual nodes will use the same versions as the ones that were used to submit the jobs
3) when running standalone SPM (only "full batteries" SPM standalone version, which includes popular toolboxes with CONN among them; this will be released some time soon by the SPM group); submitted jobs will use calls to "spm12" or "run_spm12.sh" executables (the first one found in your system's path); if there are multiple standalone-SPM or MCR versions in your cluster, individual nodes will use the same versions as the ones that were used to submit the jobs
@Damon: Let me know if this answers your questions or if I am misinterpreting (perhaps you were suggesting that it might be helpful to let the user decide whether submitted jobs will use matlab or standalone cluster implementations, or something along these lines?)
Last, regarding changes across versions, the file change.log in each release contains a detailed list of all changes in each version (the ones marked with "*" are the main ones that you want to pay special attention to in case you are updating in the middle of some already-existing CONN project)
Hope this helps
Alfonso
Originally posted by Damon Lamb:
Good afternoon All,
To piggyback on Eric's query, it would be very helpful to be able to use MCR instead of */bin/matlab to run on our cluster. So, regardless of the GUI/local processing being standalone, a standalone non-interactive build for parallel jobs on the cluster would be most useful.
Thank you,
Damon
Originally posted by eric Porges:
To piggyback on Eric's query, it would be very helpful to be able to use MCR instead of */bin/matlab to run on our cluster. So, regardless of the GUI/local processing being standalone, a standalone non-interactive build for parallel jobs on the cluster would be most useful.
Thank you,
Damon
Originally posted by eric Porges:
Alfonso and the CONN team, thank you very much
for all of your excellent work!
Couple quick questions.
1. Are there plans to have an updated version standalone CONN releases soon? It appears 15g was the last version.
2. Is there a table explaining substantive changes across versions? At the moment, we are planning to use your preprocessing functionality prior to a conventional analysis in SPM. Our dataset is very large and our cluster has a limited number of Matlab licenses that can be checked out at a given time, so the standalone provides a major time benefit. If the preprocessing is comparable, we would could just use the existing standalone version.
thanks again!
Couple quick questions.
1. Are there plans to have an updated version standalone CONN releases soon? It appears 15g was the last version.
2. Is there a table explaining substantive changes across versions? At the moment, we are planning to use your preprocessing functionality prior to a conventional analysis in SPM. Our dataset is very large and our cluster has a limited number of Matlab licenses that can be checked out at a given time, so the standalone provides a major time benefit. If the preprocessing is comparable, we would could just use the existing standalone version.
thanks again!
Threaded View
| Title | Author | Date |
|---|---|---|
| eric Porges | Apr 18, 2017 | |
| Damon Lamb | Apr 18, 2017 | |
| Alfonso Nieto-Castanon | Apr 18, 2017 | |
| Damon Lamb | May 4, 2017 | |