Platform LSF uses job
slots to specify the number of processes that are allowed to run concurrently
on a machine. A machine cannot run more concurrent processes than
it has job slots. The default number of job slots for a machine is
the same as the number of processor cores in the machine.
However, you can configure
hosts with fast processors to have more jobs slots than the number
of cores by setting the MXJ value for the given host to a fixed number
of job slots. This enables the more powerful host to execute more
jobs concurrently to take advantage of the processor’s speed.
To change the number
of job slots on a grid node, follow these steps:
-
Log on to the grid control
server as the LSF Administrator (lsfadmin).
-
Open the file
lsb.hosts,
which is located in the directory
LSF-install-dir\conf\lsbatch\cluster-name\configdir.
This is the LSF batch configuration file. Locate the Host section
of the file, which contains an entry for a
default grid
node.
Begin
Host
HOST_NAME MXJ r1m pg ls tmp DISPATCH_WINDOW #Keywords
default ! () () () () () #Example
End Host
-
Edit this file to specify
the maximum number of job slots for all nodes or for each node. There
are two ways you can do this.
-
To specify the total number of
job slots per node, edit the line for the
default node.
Here is an example:
Begin Host
HOST_NAME MXJ r1m pg ls tmp DISPATCH_WINDOW #Keywords
default ! () () () () () #Example
End Host
The value ! represents
one job slot per core for each node in the grid. You can replace this
value with a number that specifies the maximum number of job slots
on each node, regardless of the number of cores. For example, a value
of ! on a machine with 16 cores results in 16 job slots, while a value
of 2 on a machine with 16 cores results in just 2 job slots.
-
To specify the total number of
jobs slots per node, add a line for each node in the grid. Here is
an example:
Begin
Host
HOST_NAME MXJ r1m pg ls tmp DISPATCH_WINDOW #Keywords
default ! () () () () () #Example
D1234 16 () () () () () #Example
D1235 16 () () () () () #Example
D1236 16 () () () () () #Example
D1237 16 () () () () () #Example
D1238 16 () () () () () #Example
End Host
Each line designates
the concurrent execution of 16 jobs on each node.
-
-
Verify the LSF batch
configuration file by entering this command at the command prompt:
badmin
reconfig
-
For details about using
this command, see
Platform LSF Reference.