<param type="string" name="separationaxis" ignorecase="true" subparam="true"><shortdescription>Axis to do parallelization across(scan, spw, baseline, auto)</shortdescription><description>Axis to do parallelization across(scan, spw, baseline, auto)</description>
<casaxml xsi:schemaLocation="http://casa.nrao.edu/schema/casa.xsd file:///opt/casa/code/xmlcasa/xml/casa.xsd" xmlns="http://casa.nrao.edu/schema/psetTypes.html" xmlns:xsi="http://www.w3.org/2001/XMLSchema-instance">
<task type="function" name="partition" category="manipulation">
<shortdescription>Task to produce Multi-MSs using parallelism</shortdescription>
Partition is a task to create a Multi-MS out of an MS. General selection
parameters are included, and one or all of the various data columns
(DATA, LAG_DATA and/or FLOAT_DATA, and possibly MODEL_DATA and/or
CORRECTED_DATA) can be selected.
The partition task creates a Multi-MS in parallel, using the CASA MPI framework.
The user should start CASA as follows in order to run it in parallel.
1) Start CASA on a single node with 8 engines. The first engine will be used as the
MPIClient, where the user will see the CASA prompt. All other engines will be used
as MPIServers and will process the data in parallel.
mpicasa -n 8 casa --nogui --log2term
2) Running on a group of nodes in a cluster.
mpicasa -hostfile user_hostfile casa ....
where user_hostfile contains the names of the nodes and the number of engines to use
in each one of them. Example:
If CASA is started without mpicasa, it is still possible to create an MMS, but
the processing will be done in sequential.
A multi-MS is structured to have a reference MS on the top directory and a
sub-directory called SUBMSS, which contain each partitioned sub-MS. The
reference MS contains links to the sub-tables of the first sub-MS. The other
sub-MSs contain a copy of the sub-tables each. A multi-MS looks like this in disk.
ANTENNA FLAG_CMD POLARIZATION SPECTRAL_WINDOW table.dat
DATA_DESCRIPTION HISTORY PROCESSOR STATE table.info
FEED OBSERVATION SORTED_TABLE SUBMSS WEATHER
FIELD POINTING SOURCE SYSCAL
ngc5921.0000.ms/ ngc5921.0002.ms/ ngc5921.0004.ms/ ngc5921.0006.ms/
ngc5921.0001.ms/ ngc5921.0003.ms/ ngc5921.0005.ms/
Inside casapy, one can use the task listpartition to list the information
When partition processes an MMS in parallel, each sub-MS is processed independently in an engine.
The log messages of the engines are identified by the string MPIServer-#, where # gives the number
of the engine running that process. When the task runs sequentially, it shows the MPIClient text
in the origin of the log messages or does not show anything.
<param type="path" name="vis" mustexist="true"><shortdescription>Name of input measurement set</shortdescription><description>Name of input measurement set</description>