Differences

This shows you the differences between two versions of the page.

Link to this comparison view

Both sides previous revision Previous revision
faq:what_are_the_different_approaches_i_can_take_for_distributed_computing [2018/03/02 10:13]
robert [FieldTrip qsub toolbox]
faq:what_are_the_different_approaches_i_can_take_for_distributed_computing [2018/03/02 10:17] (current)
robert
Line 34: Line 34:
 If you also have the [[http://​www.mathworks.nl/​products/​distriben|distributed computing engines]], which have to be installed on a compute cluster, you can use those as remote engines for executing the separate jobs. If you also have the [[http://​www.mathworks.nl/​products/​distriben|distributed computing engines]], which have to be installed on a compute cluster, you can use those as remote engines for executing the separate jobs.
  
-==== FieldTrip ​peer computing ​toolbox ====+==== FieldTrip ​qsub toolbox ====
  
-The FieldTrip peer computing toolbox was developed to harness ​the computational resources in an unorganized organization,​ e.g. to use the computers ​of your office mates during the night, or to use multiple linux computers that are not clustered. More documentation on peer computing ​is found [[development:peer|here]].+At the Donders we have a Linux compute cluster that is managed with a batch queueing system that allows users to submit large batches ​of jobs to run in parallel. The Donders cluster runs Torque/​Maui,​ but other PBS systems, SLURM, LSF, Sun Grid Engine and Oracle Grid Engine ​are also supported. More documentation on the qsub toolbox ​is found [[/development/qsub|here]].
  
-The syntax you would use to distribute jobs with the peer system is identitcal ​to DFEVAL, i.e.+To facilitate distributed computing on the compute cluster, we have implemented a qsub wrapper function within MATLAB, which for the end-user works similar ​to DFEVAL ​and PEERCELLFUN.
  
   for i=1:Nsubj   for i=1:Nsubj
Line 44: Line 44:
     cfg{i}.dataset = sprintf('​subject%d.eeg'​);​     cfg{i}.dataset = sprintf('​subject%d.eeg'​);​
   end   end
-  rawdata = peercellfun(@ft_preprocessing,​ cfg);+  rawdata = qsubcellfun(@ft_preprocessing,​ cfg);
  
-See **[[/​reference/​peercellfun]]**, **[[/​reference/​peerfeval]]** and **[[/​reference/​peerget]]** for details.+See **[[/​reference/​qsubcellfun]]**, **[[/​reference/​qsubfeval]]** and **[[/​reference/​qsubget]]** for details.
  
-==== FieldTrip ​qsub toolbox ====+==== FieldTrip ​peer computing ​toolbox ====
  
-At the Donders we have a Linux compute cluster that is managed with a batch queueing system that allows users to submit large batches of jobs to run in parallelThe Donders cluster runs Torque/Mauibut other PBS systems, SLURM, LSF, Sun Grid Engine and Oracle Grid Engine ​are also supported. More documentation on the qsub toolbox ​is found [[/​development/​qsub|here]].+The FieldTrip peer computing toolbox was developed ​to harness the computational resources ​in an unorganized organization,​ e.g. to use the computers of your office mates during the nightor to use multiple linux computers that are not clustered. More documentation on the peer toolbox ​can be found [[/​development/​peer|here]].
  
-To facilitate distributed computing on the compute cluster, we have implemented a qsub wrapper function within MATLAB, which for the end-user works similar ​to DFEVAL ​and PEERCELLFUN.+<note warning>​Please note that this requires compilation of some mex files. At this moment it is not actively supported.</​note>​ 
 + 
 +The syntax you would use to distribute jobs with the peer system is identitcal ​to DFEVAL, i.e.
  
   for i=1:Nsubj   for i=1:Nsubj
Line 58: Line 60:
     cfg{i}.dataset = sprintf('​subject%d.eeg'​);​     cfg{i}.dataset = sprintf('​subject%d.eeg'​);​
   end   end
-  rawdata = qsubcellfun(@ft_preprocessing,​ cfg);+  rawdata = peercellfun(@ft_preprocessing,​ cfg); 
 + 
 +See **[[/​reference/​peercellfun]]**,​ **[[/​reference/​peerfeval]]** and **[[/​reference/​peerget]]** for details. 
 + 
 +==== FieldTrip engine toolbox ==== 
 + 
 +The FieldTrip-engine toolbox was designed for distributed computing on a massive multicore computer without requiring the MATLAB parallel toolbox. More documentation on the engine toolbox is found [[/​development/​engine|here]]. 
 + 
 +<note warning>​Please note that this requires compilation of some mex files. At this moment it is not actively supported.</​note>​ 
  
-See **[[/​reference/​qsubcellfun]]**,​ **[[/​reference/​qsubfeval]]** and **[[/​reference/​qsubget]]** for details.