Skip to content
Snippets Groups Projects
Commit 5ecc5cf3 authored by Morris Jette's avatar Morris Jette
Browse files

Update to Cray docs

parent bdf26102
No related branches found
No related tags found
No related merge requests found
...@@ -489,46 +489,8 @@ ulimit -t unlimited # max amount of CPU time in seconds ...@@ -489,46 +489,8 @@ ulimit -t unlimited # max amount of CPU time in seconds
ulimit -d unlimited # max size of a process's data segment in KB ulimit -d unlimited # max size of a process's data segment in KB
</pre> </pre>
<a name="gang"><h3>Gang Scheduling</h3></a>
<p>SLURM version 2.3.0-pre7 (or higher) can be used with BASIL version 1.2 or
higher to gang schedule (time-slice) parallell jobs. Review SLURM's
<a href="gang_scheduling.html">Gang Scheduling</a> guide for general
information about gang scheduling. Be certain to configure approriate values
for default and/or maximum memory limits (e.g. <b>DefMemPerNode</b> and
<b>MaxMemPerNode</b>), enable enforcement of memory limits (e.g.
<b>SelectTypeParameters=CR_Memory</b>), set an appropriate multi-programming
level in each partition (e.g. <b>Shared=FORCE:2</b>), and enable gang
scheduling (e.g. <b>PreemptMode=GANG</b>).</p>
<pre>
# Excerpt of slurm.conf
DefMemPerNode=8000
MaxMemPerNode=8000
SelectType=select/cray
SelectTypeParameters=CR_Memeory
PreemptMode=GANG
FrontEndName=frontend[00-03]
NodeName=nid[00000-00511] Memory=32000 CPUs=6
PartitionName=batch Nodes=nid[00000-00511] Shared=FORCE:4
</pre>
<p>ALPS' configuration file, <i>alps.conf</i>, also requires some modifcations
in order to support gang scheduling. A sample configuration is shown below.
Of particular note, <b>slice</b> is the time slice interval and should be set
equal to zero in order for an entity external to ALPS (i.e. SLURM) to control
when jobs are suspended and resumed. <b>loadLimit</b> indicates the maximum
number of jobs per node and should be set to 2 for a Gemini interconnect and
up to 4 otherwise.</p>
<pre>
# Excerpt of alps.conf
nodeShare share # over-subscribe nodes
sliceArch xt
slice 0
loadLimit 4 # set to 2 for Gemini
</pre>
<p class="footer"><a href="#top">top</a></p> <p class="footer"><a href="#top">top</a></p>
<p style="text-align:center;">Last modified 28 June 2011</p></td> <p style="text-align:center;">Last modified 11 July 2011</p></td>
<!--#include virtual="footer.txt"--> <!--#include virtual="footer.txt"-->
...@@ -48,8 +48,7 @@ gang scheduling: ...@@ -48,8 +48,7 @@ gang scheduling:
<LI> <LI>
<B>SelectType</B>: The SLURM gang scheduler supports nodes <B>SelectType</B>: The SLURM gang scheduler supports nodes
allocated by the <I>select/linear</I> plugin and socket/core/CPU resources allocated by the <I>select/linear</I> plugin and socket/core/CPU resources
allocated by the <I>select/cons_res</I> plugin. The <I>select/cray</I> is allocated by the <I>select/cons_res</I> plugin.
also supported for Cray computers running BASIL version 1.2 or higher.
</LI> </LI>
<LI> <LI>
<B>SelectTypeParameter</B>: Since resources will be getting overallocated <B>SelectTypeParameter</B>: Since resources will be getting overallocated
......
...@@ -1472,6 +1472,10 @@ See the partition \fBShared\fR parameter for more information. ...@@ -1472,6 +1472,10 @@ See the partition \fBShared\fR parameter for more information.
\fBselect/bluegene\fR \fBselect/bluegene\fR
for a three\-dimensional BlueGene system. for a three\-dimensional BlueGene system.
The default value is "select/bluegene" for BlueGene systems. The default value is "select/bluegene" for BlueGene systems.
.TP
\fBselect/cray\fR
for a Cray system.
The default value is "select/cray" for all Cray systems.
.RE .RE
.TP .TP
......
0% Loading or .
You are about to add 0 people to the discussion. Proceed with caution.
Finish editing this message first!
Please register or to comment