Data partition

partition task highlights

Data partitioning is the process by which a given, existing MeasurementSet is partitioned into either several individual MeasurementSets, or a "Multi-MS". A Multi-MS is a CASA construct designed to facilitate parallel processing; essentially a Multi-MS or MMS comprises an internally partitioned MeasurementSet.  Partitioning can be done with the partition task; as for many other MS transforming or manipulating tasks, partition is nothing but a simplified front-end to mstransform. Although mstransform can in fact do the same operations it is strongly recommended to use partition for partitioning operations, and in particular for creation of Multi-MS's.

 

The key parameters that control creation of an MMS are shown below:

CASA <1>: inp mstransform


createmms           =       True        #  Should this create a multi-MS output
     separationaxis =     'auto'        #  Axis to do parallelization across(scan, spw, baseline, auto)
     numsubms       =     'auto'        #  The number of SubMSs to create (auto or any number)

 

  • separationaxis: Specifies the partition axis, across scan, spw, baseline, or auto. The default mode is auto, which parts the data across both scan and spw distributing the data of each scan and spw across all Sub-MS in the most balanced way possible.
  • numsubms: When this parameter is assigned to an integer it refers to the number of sub-MSs that the output MMS would have. The default is auto which means producing as many Sub-MSs as parallel cores.

 

For more details please refer to the Parallel Processing chapter.