[#8140] Parallel bedpostx processing

Please log in

State: more information
Open
Date:
2016-08-01 05:03
Priority: more information
3
Submitted By:
Richard Morris (muswellbrook)
Assigned To: more information
Nobody (None)
Summary: more information
Parallel bedpostx processing

Detailed description

Dear support, 

I am a new user of the NITRC-CE environment on AWS, and FSL. I would like to know whether parallel processing (SGE) is enabled in the NITRC-CE? When I run FSLPARALLEL=1 bedpostx, I can see that only two virtual cores are utilised, even on a 4 or 8 vCPU AMI. I expected to see more cores used by the parallel pool.

Thanks for any help you can offer,

Rich

Response

Message

Date: 2016-08-02 05:06
Sender: David Kennedy

In addition, the NITRC-CE version should be in the footer of the NITRC-CE web interface.


Date: 2016-08-02 04:24
Sender: David Kennedy

Also, can you tell me which version of the NITRC-CE you are using?

In the EC2 console, for a given instance, there should be an entry like:
     AMI ID: NITRC-CE v.42 (ami-903001fa)

Thanks.
DK


Date: 2016-08-02 04:11
Sender: David Kennedy

I think you're on the right track (given the qmon question in the forum). The expected behavior for the NITRC-CE when they are started is to default the number of processors to the number of cores, and the number of slots to 2x number_of_cores in the SGE setup.

I will have to review if these SGE settings follow when the instance types get changed, or only automatically match the instance configuration at initial startup. It is common to start with a small (or micro) instance to get everything setup, then switch to something bigger (xlarge, or 4xlarge, etc.) for more cores when ready to run.

DK

Attached Files:

Name Download
No Files Currently Attached

Changes:

Field Old Value Date By
New Message2016-08-02 05:06dnkennedy
New Message2016-08-02 04:24dnkennedy
New Message2016-08-02 04:11dnkennedy