Advanced MPI scheduling: Difference between revisions

Advanced MPI scheduling (view source)

Revision as of 17:09, 12 April 2019

457 bytes removed , 5 years ago

remove redundant table of node memory sizes

Rdickson

Bureaucrats, cc_docs_admin, cc_staff

2,879

edits

@@ Line 111: / Line 111: @@
 <!--T:30-->
-Note however that if you need more memory per node than the smallest node provides (e.g. more than 125 GiB at Graham) then you should not use <code>--mem=0</code>, but request the amount explicitly. Furthermore, some memory on each node is reserved for the operating system, so the largest amount your job can request and still qualify for each node type is as follows:
+If you need more memory per node than the smallest node provides (e.g. more than 125 GiB at Graham) then you should not use <code>--mem=0</code>, but request the amount explicitly. Furthermore, some memory on each node is reserved for the operating system.  To find the largest amount your job can request and still qualify for a given node type, consult the "Available memory" column of the "Node types and characteristics" table on cluster description page:
+* [[Béluga/en#Node_types_and_characteristics|Béluga node types and characteristics]]
-<!--T:31-->
+* [[Cedar#Node_types_and_characteristics|Cedar node types and characteristics]]
-{| class="wikitable"
+* [[Graham#Node_types_and_characteristics|Graham node types and characteristics]]
-|-
-! Nominal node size  !! Maximum Slurm request  !! Where?
-|-
-| 128 GB = 125 GiB   || --mem=128000M          || Graham, Cedar Broadwell
-|-
-| 192 GB = 187 GiB   || --mem=192000M          || Cedar Skylake
-|-
-| 256 GB =           || --mem=256000M          || Graham, Cedar Broadwell
-|-
-| 512 GB =           || --mem=512000M          || Graham, Cedar Broadwell
-|-
-| 1.5 TB =           || --mem=1500G            || Cedar Broadwell
-|-
-| 3 TB =             || --mem=3000G            || Graham, Cedar Broadwell
-|}
-<!--T:32-->
-Requesting <code>--mem=256G</code> instead of <code>--mem=256000M</code> means the job will not schedule on a so-called 256GB node. The job will not be rejected by Slurm, it will just wait longer that it needs to for a rarer, larger node.
 ==== Few cores, single node ==== <!--T:6-->