Hi David,
there is a new robust motion correction used in 5.0 which is more accurate than the flirt, but it needs a lot of memory.
You can go back to the old version by adding the parameter: -flirt-motioncor
Best, Martin
On Tue, 2010-08-31 at 18:01 -0700, David Mischel wrote:
We run our Freesurfer (recon-all) jobs on a grid (Sun Grid Engine) which uses a single NFS file server for all data. Since we have begun using Freesurfer 5.0.0 we have noticed very intense I/O activity as soon as recon-all jobs begin. The level of activity is markedly greater than we have seen in previous versions. The effective cumulative impact on our file server, when scores of jobs are launched at once, is significant and degrades the server's response.
Has something changed about how data files are accessed at the beginning of a recon-all job? If so, is there a way to "ease into it" so that the initial activity is more distributed as processing continues?
< david
David Mischel
Manager of IT
Center for Imaging of Neurodegenerative Diseases (CIND)
http://www.cind.research.va.gov/
VA Medical Center
4150 Clement Street, 114M
San Francisco, CA 94121
voice: 415-221-4810 x3864
fax: 415-668-2864
Freesurfer mailing list Freesurfer@nmr.mgh.harvard.edu https://mail.nmr.mgh.harvard.edu/mailman/listinfo/freesurfer
I think David was talking about more disk I/O. Was that increased also? On Wed, 1 Sep 2010, Martin Reuter wrote:
Hi David,
there is a new robust motion correction used in 5.0 which is more accurate than the flirt, but it needs a lot of memory.
You can go back to the old version by adding the parameter: -flirt-motioncor
Best, Martin
On Tue, 2010-08-31 at 18:01 -0700, David Mischel wrote:
We run our Freesurfer (recon-all) jobs on a grid (Sun Grid Engine) which uses a single NFS file server for all data. Since we have begun using Freesurfer 5.0.0 we have noticed very intense I/O activity as soon as recon-all jobs begin. The level of activity is markedly greater than we have seen in previous versions. The effective cumulative impact on our file server, when scores of jobs are launched at once, is significant and degrades the server's response.
Has something changed about how data files are accessed at the beginning of a recon-all job? If so, is there a way to "ease into it" so that the initial activity is more distributed as processing continues?
< david
David Mischel
Manager of IT
Center for Imaging of Neurodegenerative Diseases (CIND)
http://www.cind.research.va.gov/
VA Medical Center
4150 Clement Street, 114M
San Francisco, CA 94121
voice: 415-221-4810 x3864
fax: 415-668-2864
Freesurfer mailing list Freesurfer@nmr.mgh.harvard.edu https://mail.nmr.mgh.harvard.edu/mailman/listinfo/freesurfer
Freesurfer mailing list Freesurfer@nmr.mgh.harvard.edu https://mail.nmr.mgh.harvard.edu/mailman/listinfo/freesurfer
Hm, not sure, but I don't see a reason why there should be more IO in the motion correction step. It reads the inputs and writes the rawavg as before.
Maybe Nick knows what could cause this?
-Martin
On Wed, 2010-09-01 at 11:40 -0400, Bruce Fischl wrote:
I think David was talking about more disk I/O. Was that increased also? On Wed, 1 Sep 2010, Martin Reuter wrote:
Hi David,
there is a new robust motion correction used in 5.0 which is more accurate than the flirt, but it needs a lot of memory.
You can go back to the old version by adding the parameter: -flirt-motioncor
Best, Martin
On Tue, 2010-08-31 at 18:01 -0700, David Mischel wrote:
We run our Freesurfer (recon-all) jobs on a grid (Sun Grid Engine) which uses a single NFS file server for all data. Since we have begun using Freesurfer 5.0.0 we have noticed very intense I/O activity as soon as recon-all jobs begin. The level of activity is markedly greater than we have seen in previous versions. The effective cumulative impact on our file server, when scores of jobs are launched at once, is significant and degrades the server's response.
Has something changed about how data files are accessed at the beginning of a recon-all job? If so, is there a way to "ease into it" so that the initial activity is more distributed as processing continues?
< david
David Mischel
Manager of IT
Center for Imaging of Neurodegenerative Diseases (CIND)
http://www.cind.research.va.gov/
VA Medical Center
4150 Clement Street, 114M
San Francisco, CA 94121
voice: 415-221-4810 x3864
fax: 415-668-2864
Freesurfer mailing list Freesurfer@nmr.mgh.harvard.edu https://mail.nmr.mgh.harvard.edu/mailman/listinfo/freesurfer
Freesurfer mailing list Freesurfer@nmr.mgh.harvard.edu https://mail.nmr.mgh.harvard.edu/mailman/listinfo/freesurfer
freesurfer@nmr.mgh.harvard.edu
-
Bruce Fischl -
Martin Reuter